Mercurial > cortex

annotate org/vision.org @ 262:0e85237d27a7

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Added a diagram explaining RenderManager
author Dylan Holmes <ocsenave@gmail.com>
date Mon, 13 Feb 2012 06:48:40 -0600
parents 02b2e6f3fb43
children f8227f6d4ac6
rev   line source
rlm@34 1 #+title: Simulated Sense of Sight
rlm@23 2 #+author: Robert McIntyre
rlm@23 3 #+email: rlm@mit.edu
rlm@38 4 #+description: Simulated sight for AI research using JMonkeyEngine3 and clojure
rlm@34 5 #+keywords: computer vision, jMonkeyEngine3, clojure
rlm@23 6 #+SETUPFILE: ../../aurellem/org/setup.org
rlm@23 7 #+INCLUDE: ../../aurellem/org/level-0.org
rlm@23 8 #+babel: :mkdirp yes :noweb yes :exports both
rlm@23 9
rlm@194 10 * Vision
rlm@216 11
rlm@212 12 Vision is one of the most important senses for humans, so I need to
rlm@212 13 build a simulated sense of vision for my AI. I will do this with
rlm@212 14 simulated eyes. Each eye can be independely moved and should see its
rlm@212 15 own version of the world depending on where it is.
rlm@212 16
rlm@218 17 Making these simulated eyes a reality is simple bacause jMonkeyEngine
rlm@218 18 already conatains extensive support for multiple views of the same 3D
rlm@218 19 simulated world. The reason jMonkeyEngine has this support is because
rlm@218 20 the support is necessary to create games with split-screen
rlm@218 21 views. Multiple views are also used to create efficient
rlm@212 22 pseudo-reflections by rendering the scene from a certain perspective
rlm@212 23 and then projecting it back onto a surface in the 3D world.
rlm@212 24
rlm@218 25 #+caption: jMonkeyEngine supports multiple views to enable split-screen games, like GoldenEye, which was one of the first games to use split-screen views.
rlm@212 26 [[../images/goldeneye-4-player.png]]
rlm@212 27
rlm@213 28 * Brief Description of jMonkeyEngine's Rendering Pipeline
rlm@212 29
rlm@213 30 jMonkeyEngine allows you to create a =ViewPort=, which represents a
rlm@213 31 view of the simulated world. You can create as many of these as you
rlm@213 32 want. Every frame, the =RenderManager= iterates through each
rlm@213 33 =ViewPort=, rendering the scene in the GPU. For each =ViewPort= there
rlm@213 34 is a =FrameBuffer= which represents the rendered image in the GPU.
rlm@151 35
ocsenave@262 36 #+caption: =ViewPorts= are cameras in the world. During each frame, the =Rendermanager= records a snapshot of what each view is currently seeing.
ocsenave@262 37 #+attr_html:width="400"
ocsenave@262 38 [[../images/diagram_rendermanager.png]]
ocsenave@262 39
rlm@213 40 Each =ViewPort= can have any number of attached =SceneProcessor=
rlm@213 41 objects, which are called every time a new frame is rendered. A
rlm@219 42 =SceneProcessor= recieves its =ViewPort's= =FrameBuffer= and can do
rlm@219 43 whatever it wants to the data. Often this consists of invoking GPU
rlm@219 44 specific operations on the rendered image. The =SceneProcessor= can
rlm@219 45 also copy the GPU image data to RAM and process it with the CPU.
rlm@151 46
rlm@213 47 * The Vision Pipeline
rlm@151 48
rlm@213 49 Each eye in the simulated creature needs it's own =ViewPort= so that
rlm@213 50 it can see the world from its own perspective. To this =ViewPort=, I
rlm@214 51 add a =SceneProcessor= that feeds the visual data to any arbitray
rlm@213 52 continuation function for further processing. That continuation
rlm@213 53 function may perform both CPU and GPU operations on the data. To make
rlm@213 54 this easy for the continuation function, the =SceneProcessor=
rlm@213 55 maintains appropriatly sized buffers in RAM to hold the data. It does
rlm@218 56 not do any copying from the GPU to the CPU itself because it is a slow
rlm@218 57 operation.
rlm@214 58
rlm@213 59 #+name: pipeline-1
rlm@213 60 #+begin_src clojure
rlm@113 61 (defn vision-pipeline
rlm@34 62 "Create a SceneProcessor object which wraps a vision processing
rlm@113 63 continuation function. The continuation is a function that takes
rlm@113 64 [#^Renderer r #^FrameBuffer fb #^ByteBuffer b #^BufferedImage bi],
rlm@113 65 each of which has already been appropiately sized."
rlm@23 66 [continuation]
rlm@23 67 (let [byte-buffer (atom nil)
rlm@113 68 renderer (atom nil)
rlm@113 69 image (atom nil)]
rlm@23 70 (proxy [SceneProcessor] []
rlm@23 71 (initialize
rlm@23 72 [renderManager viewPort]
rlm@23 73 (let [cam (.getCamera viewPort)
rlm@23 74 width (.getWidth cam)
rlm@23 75 height (.getHeight cam)]
rlm@23 76 (reset! renderer (.getRenderer renderManager))
rlm@23 77 (reset! byte-buffer
rlm@23 78 (BufferUtils/createByteBuffer
rlm@113 79 (* width height 4)))
rlm@113 80 (reset! image (BufferedImage.
rlm@113 81 width height
rlm@113 82 BufferedImage/TYPE_4BYTE_ABGR))))
rlm@23 83 (isInitialized [] (not (nil? @byte-buffer)))
rlm@23 84 (reshape [_ _ _])
rlm@23 85 (preFrame [_])
rlm@23 86 (postQueue [_])
rlm@23 87 (postFrame
rlm@23 88 [#^FrameBuffer fb]
rlm@23 89 (.clear @byte-buffer)
rlm@113 90 (continuation @renderer fb @byte-buffer @image))
rlm@23 91 (cleanup []))))
rlm@213 92 #+end_src
rlm@213 93
rlm@213 94 The continuation function given to =(vision-pipeline)= above will be
rlm@213 95 given a =Renderer= and three containers for image data. The
rlm@218 96 =FrameBuffer= references the GPU image data, but the pixel data can
rlm@218 97 not be used directly on the CPU. The =ByteBuffer= and =BufferedImage=
rlm@219 98 are initially "empty" but are sized to hold the data in the
rlm@213 99 =FrameBuffer=. I call transfering the GPU image data to the CPU
rlm@213 100 structures "mixing" the image data. I have provided three functions to
rlm@213 101 do this mixing.
rlm@213 102
rlm@213 103 #+name: pipeline-2
rlm@213 104 #+begin_src clojure
rlm@113 105 (defn frameBuffer->byteBuffer!
rlm@113 106 "Transfer the data in the graphics card (Renderer, FrameBuffer) to
rlm@113 107 the CPU (ByteBuffer)."
rlm@113 108 [#^Renderer r #^FrameBuffer fb #^ByteBuffer bb]
rlm@113 109 (.readFrameBuffer r fb bb) bb)
rlm@113 110
rlm@113 111 (defn byteBuffer->bufferedImage!
rlm@113 112 "Convert the C-style BGRA image data in the ByteBuffer bb to the AWT
rlm@113 113 style ABGR image data and place it in BufferedImage bi."
rlm@113 114 [#^ByteBuffer bb #^BufferedImage bi]
rlm@113 115 (Screenshots/convertScreenShot bb bi) bi)
rlm@113 116
rlm@113 117 (defn BufferedImage!
rlm@113 118 "Continuation which will grab the buffered image from the materials
rlm@113 119 provided by (vision-pipeline)."
rlm@113 120 [#^Renderer r #^FrameBuffer fb #^ByteBuffer bb #^BufferedImage bi]
rlm@113 121 (byteBuffer->bufferedImage!
rlm@113 122 (frameBuffer->byteBuffer! r fb bb) bi))
rlm@213 123 #+end_src
rlm@112 124
rlm@213 125 Note that it is possible to write vision processing algorithms
rlm@213 126 entirely in terms of =BufferedImage= inputs. Just compose that
rlm@213 127 =BufferedImage= algorithm with =(BufferedImage!)=. However, a vision
rlm@213 128 processing algorithm that is entirely hosted on the GPU does not have
rlm@213 129 to pay for this convienence.
rlm@213 130
rlm@214 131 * COMMENT asdasd
rlm@213 132
rlm@213 133 (vision creature) will take an optional :skip argument which will
rlm@213 134 inform the continuations in scene processor to skip the given
rlm@213 135 number of cycles 0 means that no cycles will be skipped.
rlm@213 136
rlm@213 137 (vision creature) will return [init-functions sensor-functions].
rlm@213 138 The init-functions are each single-arg functions that take the
rlm@213 139 world and register the cameras and must each be called before the
rlm@213 140 corresponding sensor-functions. Each init-function returns the
rlm@213 141 viewport for that eye which can be manipulated, saved, etc. Each
rlm@213 142 sensor-function is a thunk and will return data in the same
rlm@213 143 format as the tactile-sensor functions the structure is
rlm@213 144 [topology, sensor-data]. Internally, these sensor-functions
rlm@213 145 maintain a reference to sensor-data which is periodically updated
rlm@213 146 by the continuation function established by its init-function.
rlm@213 147 They can be queried every cycle, but their information may not
rlm@213 148 necessairly be different every cycle.
rlm@213 149
rlm@214 150 * Physical Eyes
rlm@214 151
rlm@214 152 The vision pipeline described above handles the flow of rendered
rlm@214 153 images. Now, we need simulated eyes to serve as the source of these
rlm@214 154 images.
rlm@214 155
rlm@214 156 An eye is described in blender in the same way as a joint. They are
rlm@214 157 zero dimensional empty objects with no geometry whose local coordinate
rlm@214 158 system determines the orientation of the resulting eye. All eyes are
rlm@214 159 childern of a parent node named "eyes" just as all joints have a
rlm@214 160 parent named "joints". An eye binds to the nearest physical object
rlm@214 161 with =(bind-sense=).
rlm@214 162
rlm@214 163 #+name: add-eye
rlm@214 164 #+begin_src clojure
rlm@215 165 (in-ns 'cortex.vision)
rlm@215 166
rlm@214 167 (defn add-eye!
rlm@214 168 "Create a Camera centered on the current position of 'eye which
rlm@214 169 follows the closest physical node in 'creature and sends visual
rlm@215 170 data to 'continuation. The camera will point in the X direction and
rlm@215 171 use the Z vector as up as determined by the rotation of these
rlm@215 172 vectors in blender coordinate space. Use XZY rotation for the node
rlm@215 173 in blender."
rlm@214 174 [#^Node creature #^Spatial eye]
rlm@214 175 (let [target (closest-node creature eye)
rlm@214 176 [cam-width cam-height] (eye-dimensions eye)
rlm@215 177 cam (Camera. cam-width cam-height)
rlm@215 178 rot (.getWorldRotation eye)]
rlm@214 179 (.setLocation cam (.getWorldTranslation eye))
rlm@218 180 (.lookAtDirection
rlm@218 181 cam ; this part is not a mistake and
rlm@218 182 (.mult rot Vector3f/UNIT_X) ; is consistent with using Z in
rlm@218 183 (.mult rot Vector3f/UNIT_Y)) ; blender as the UP vector.
rlm@214 184 (.setFrustumPerspective
rlm@215 185 cam 45 (/ (.getWidth cam) (.getHeight cam)) 1 1000)
rlm@215 186 (bind-sense target cam) cam))
rlm@214 187 #+end_src
rlm@214 188
rlm@214 189 Here, the camera is created based on metadata on the eye-node and
rlm@214 190 attached to the nearest physical object with =(bind-sense)=
rlm@214 191
rlm@214 192
rlm@214 193 ** The Retina
rlm@214 194
rlm@214 195 An eye is a surface (the retina) which contains many discrete sensors
rlm@218 196 to detect light. These sensors have can have different light-sensing
rlm@214 197 properties. In humans, each discrete sensor is sensitive to red,
rlm@214 198 blue, green, or gray. These different types of sensors can have
rlm@214 199 different spatial distributions along the retina. In humans, there is
rlm@214 200 a fovea in the center of the retina which has a very high density of
rlm@214 201 color sensors, and a blind spot which has no sensors at all. Sensor
rlm@219 202 density decreases in proportion to distance from the fovea.
rlm@214 203
rlm@214 204 I want to be able to model any retinal configuration, so my eye-nodes
rlm@214 205 in blender contain metadata pointing to images that describe the
rlm@214 206 percise position of the individual sensors using white pixels. The
rlm@214 207 meta-data also describes the percise sensitivity to light that the
rlm@214 208 sensors described in the image have. An eye can contain any number of
rlm@214 209 these images. For example, the metadata for an eye might look like
rlm@214 210 this:
rlm@214 211
rlm@214 212 #+begin_src clojure
rlm@214 213 {0xFF0000 "Models/test-creature/retina-small.png"}
rlm@214 214 #+end_src
rlm@214 215
rlm@214 216 #+caption: The retinal profile image "Models/test-creature/retina-small.png". White pixels are photo-sensitive elements. The distribution of white pixels is denser in the middle and falls off at the edges and is inspired by the human retina.
rlm@214 217 [[../assets/Models/test-creature/retina-small.png]]
rlm@214 218
rlm@214 219 Together, the number 0xFF0000 and the image image above describe the
rlm@214 220 placement of red-sensitive sensory elements.
rlm@214 221
rlm@214 222 Meta-data to very crudely approximate a human eye might be something
rlm@214 223 like this:
rlm@214 224
rlm@214 225 #+begin_src clojure
rlm@214 226 (let [retinal-profile "Models/test-creature/retina-small.png"]
rlm@214 227 {0xFF0000 retinal-profile
rlm@214 228 0x00FF00 retinal-profile
rlm@214 229 0x0000FF retinal-profile
rlm@214 230 0xFFFFFF retinal-profile})
rlm@214 231 #+end_src
rlm@214 232
rlm@214 233 The numbers that serve as keys in the map determine a sensor's
rlm@214 234 relative sensitivity to the channels red, green, and blue. These
rlm@218 235 sensitivity values are packed into an integer in the order =|_|R|G|B|=
rlm@218 236 in 8-bit fields. The RGB values of a pixel in the image are added
rlm@214 237 together with these sensitivities as linear weights. Therfore,
rlm@214 238 0xFF0000 means sensitive to red only while 0xFFFFFF means sensitive to
rlm@214 239 all colors equally (gray).
rlm@214 240
rlm@214 241 For convienence I've defined a few symbols for the more common
rlm@214 242 sensitivity values.
rlm@214 243
rlm@214 244 #+name: sensitivity
rlm@214 245 #+begin_src clojure
rlm@214 246 (defvar sensitivity-presets
rlm@214 247 {:all 0xFFFFFF
rlm@214 248 :red 0xFF0000
rlm@214 249 :blue 0x0000FF
rlm@214 250 :green 0x00FF00}
rlm@214 251 "Retinal sensitivity presets for sensors that extract one channel
rlm@219 252 (:red :blue :green) or average all channels (:all)")
rlm@214 253 #+end_src
rlm@214 254
rlm@214 255 ** Metadata Processing
rlm@214 256
rlm@214 257 =(retina-sensor-profile)= extracts a map from the eye-node in the same
rlm@214 258 format as the example maps above. =(eye-dimensions)= finds the
rlm@219 259 dimensions of the smallest image required to contain all the retinal
rlm@214 260 sensor maps.
rlm@214 261
rlm@216 262 #+name: retina
rlm@214 263 #+begin_src clojure
rlm@214 264 (defn retina-sensor-profile
rlm@214 265 "Return a map of pixel sensitivity numbers to BufferedImages
rlm@214 266 describing the distribution of light-sensitive components of this
rlm@214 267 eye. :red, :green, :blue, :gray are already defined as extracting
rlm@214 268 the red, green, blue, and average components respectively."
rlm@214 269 [#^Spatial eye]
rlm@214 270 (if-let [eye-map (meta-data eye "eye")]
rlm@214 271 (map-vals
rlm@214 272 load-image
rlm@214 273 (eval (read-string eye-map)))))
rlm@214 274
rlm@218 275 (defn eye-dimensions
rlm@218 276 "Returns [width, height] determined by the metadata of the eye."
rlm@214 277 [#^Spatial eye]
rlm@214 278 (let [dimensions
rlm@214 279 (map #(vector (.getWidth %) (.getHeight %))
rlm@214 280 (vals (retina-sensor-profile eye)))]
rlm@214 281 [(apply max (map first dimensions))
rlm@214 282 (apply max (map second dimensions))]))
rlm@214 283 #+end_src
rlm@214 284
rlm@214 285 * Eye Creation
rlm@214 286 First off, get the children of the "eyes" empty node to find all the
rlm@214 287 eyes the creature has.
rlm@216 288 #+name: eye-node
rlm@214 289 #+begin_src clojure
rlm@214 290 (defvar
rlm@214 291 ^{:arglists '([creature])}
rlm@214 292 eyes
rlm@214 293 (sense-nodes "eyes")
rlm@214 294 "Return the children of the creature's \"eyes\" node.")
rlm@214 295 #+end_src
rlm@214 296
rlm@215 297 Then, add the camera created by =(add-eye!)= to the simulation by
rlm@215 298 creating a new viewport.
rlm@214 299
rlm@216 300 #+name: add-camera
rlm@213 301 #+begin_src clojure
rlm@169 302 (defn add-camera!
rlm@169 303 "Add a camera to the world, calling continuation on every frame
rlm@34 304 produced."
rlm@167 305 [#^Application world camera continuation]
rlm@23 306 (let [width (.getWidth camera)
rlm@23 307 height (.getHeight camera)
rlm@23 308 render-manager (.getRenderManager world)
rlm@23 309 viewport (.createMainView render-manager "eye-view" camera)]
rlm@23 310 (doto viewport
rlm@23 311 (.setClearFlags true true true)
rlm@112 312 (.setBackgroundColor ColorRGBA/Black)
rlm@113 313 (.addProcessor (vision-pipeline continuation))
rlm@23 314 (.attachScene (.getRootNode world)))))
rlm@215 315 #+end_src
rlm@151 316
rlm@151 317
rlm@218 318 The eye's continuation function should register the viewport with the
rlm@218 319 simulation the first time it is called, use the CPU to extract the
rlm@215 320 appropriate pixels from the rendered image and weight them by each
rlm@218 321 sensor's sensitivity. I have the option to do this processing in
rlm@218 322 native code for a slight gain in speed. I could also do it in the GPU
rlm@218 323 for a massive gain in speed. =(vision-kernel)= generates a list of
rlm@218 324 such continuation functions, one for each channel of the eye.
rlm@151 325
rlm@216 326 #+name: kernel
rlm@215 327 #+begin_src clojure
rlm@215 328 (in-ns 'cortex.vision)
rlm@151 329
rlm@215 330 (defrecord attached-viewport [vision-fn viewport-fn]
rlm@215 331 clojure.lang.IFn
rlm@215 332 (invoke [this world] (vision-fn world))
rlm@215 333 (applyTo [this args] (apply vision-fn args)))
rlm@151 334
rlm@216 335 (defn pixel-sense [sensitivity pixel]
rlm@216 336 (let [s-r (bit-shift-right (bit-and 0xFF0000 sensitivity) 16)
rlm@216 337 s-g (bit-shift-right (bit-and 0x00FF00 sensitivity) 8)
rlm@216 338 s-b (bit-and 0x0000FF sensitivity)
rlm@216 339
rlm@216 340 p-r (bit-shift-right (bit-and 0xFF0000 pixel) 16)
rlm@216 341 p-g (bit-shift-right (bit-and 0x00FF00 pixel) 8)
rlm@216 342 p-b (bit-and 0x0000FF pixel)
rlm@216 343
rlm@216 344 total-sensitivity (* 255 (+ s-r s-g s-b))]
rlm@216 345 (float (/ (+ (* s-r p-r)
rlm@216 346 (* s-g p-g)
rlm@216 347 (* s-b p-b))
rlm@216 348 total-sensitivity))))
rlm@216 349
rlm@215 350 (defn vision-kernel
rlm@171 351 "Returns a list of functions, each of which will return a color
rlm@171 352 channel's worth of visual information when called inside a running
rlm@171 353 simulation."
rlm@151 354 [#^Node creature #^Spatial eye & {skip :skip :or {skip 0}}]
rlm@169 355 (let [retinal-map (retina-sensor-profile eye)
rlm@169 356 camera (add-eye! creature eye)
rlm@151 357 vision-image
rlm@151 358 (atom
rlm@151 359 (BufferedImage. (.getWidth camera)
rlm@151 360 (.getHeight camera)
rlm@170 361 BufferedImage/TYPE_BYTE_BINARY))
rlm@170 362 register-eye!
rlm@170 363 (runonce
rlm@170 364 (fn [world]
rlm@170 365 (add-camera!
rlm@170 366 world camera
rlm@170 367 (let [counter (atom 0)]
rlm@170 368 (fn [r fb bb bi]
rlm@170 369 (if (zero? (rem (swap! counter inc) (inc skip)))
rlm@170 370 (reset! vision-image
rlm@170 371 (BufferedImage! r fb bb bi))))))))]
rlm@151 372 (vec
rlm@151 373 (map
rlm@151 374 (fn [[key image]]
rlm@151 375 (let [whites (white-coordinates image)
rlm@151 376 topology (vec (collapse whites))
rlm@216 377 sensitivity (sensitivity-presets key key)]
rlm@215 378 (attached-viewport.
rlm@215 379 (fn [world]
rlm@215 380 (register-eye! world)
rlm@215 381 (vector
rlm@215 382 topology
rlm@215 383 (vec
rlm@215 384 (for [[x y] whites]
rlm@216 385 (pixel-sense
rlm@216 386 sensitivity
rlm@216 387 (.getRGB @vision-image x y))))))
rlm@215 388 register-eye!)))
rlm@215 389 retinal-map))))
rlm@151 390
rlm@215 391 (defn gen-fix-display
rlm@215 392 "Create a function to call to restore a simulation's display when it
rlm@215 393 is disrupted by a Viewport."
rlm@215 394 []
rlm@215 395 (runonce
rlm@215 396 (fn [world]
rlm@215 397 (add-camera! world (.getCamera world) no-op))))
rlm@215 398 #+end_src
rlm@170 399
rlm@215 400 Note that since each of the functions generated by =(vision-kernel)=
rlm@215 401 shares the same =(register-eye!)= function, the eye will be registered
rlm@215 402 only once the first time any of the functions from the list returned
rlm@215 403 by =(vision-kernel)= is called. Each of the functions returned by
rlm@215 404 =(vision-kernel)= also allows access to the =Viewport= through which
rlm@215 405 it recieves images.
rlm@215 406
rlm@215 407 The in-game display can be disrupted by all the viewports that the
rlm@215 408 functions greated by =(vision-kernel)= add. This doesn't affect the
rlm@215 409 simulation or the simulated senses, but can be annoying.
rlm@215 410 =(gen-fix-display)= restores the in-simulation display.
rlm@215 411
rlm@215 412 ** Vision!
rlm@215 413
rlm@218 414 All the hard work has been done; all that remains is to apply
rlm@215 415 =(vision-kernel)= to each eye in the creature and gather the results
rlm@215 416 into one list of functions.
rlm@215 417
rlm@216 418 #+name: main
rlm@215 419 #+begin_src clojure
rlm@170 420 (defn vision!
rlm@170 421 "Returns a function which returns visual sensory data when called
rlm@218 422 inside a running simulation."
rlm@151 423 [#^Node creature & {skip :skip :or {skip 0}}]
rlm@151 424 (reduce
rlm@170 425 concat
rlm@167 426 (for [eye (eyes creature)]
rlm@215 427 (vision-kernel creature eye))))
rlm@215 428 #+end_src
rlm@151 429
rlm@215 430 ** Visualization of Vision
rlm@215 431
rlm@215 432 It's vital to have a visual representation for each sense. Here I use
rlm@215 433 =(view-sense)= to construct a function that will create a display for
rlm@215 434 visual data.
rlm@215 435
rlm@216 436 #+name: display
rlm@215 437 #+begin_src clojure
rlm@216 438 (in-ns 'cortex.vision)
rlm@216 439
rlm@189 440 (defn view-vision
rlm@189 441 "Creates a function which accepts a list of visual sensor-data and
rlm@189 442 displays each element of the list to the screen."
rlm@189 443 []
rlm@188 444 (view-sense
rlm@188 445 (fn
rlm@188 446 [[coords sensor-data]]
rlm@188 447 (let [image (points->image coords)]
rlm@188 448 (dorun
rlm@188 449 (for [i (range (count coords))]
rlm@188 450 (.setRGB image ((coords i) 0) ((coords i) 1)
rlm@216 451 (gray (int (* 255 (sensor-data i)))))))
rlm@189 452 image))))
rlm@34 453 #+end_src
rlm@23 454
rlm@215 455 * Tests
rlm@215 456 ** Basic Test
rlm@23 457
rlm@215 458 This is a basic test for the vision system. It only tests the
rlm@215 459 vision-pipeline and does not deal with loadig eyes from a blender
rlm@215 460 file. The code creates two videos of the same rotating cube from
rlm@215 461 different angles.
rlm@23 462
rlm@215 463 #+name: test-1
rlm@23 464 #+begin_src clojure
rlm@215 465 (in-ns 'cortex.test.vision)
rlm@23 466
rlm@219 467 (defn test-pipeline
rlm@69 468 "Testing vision:
rlm@69 469 Tests the vision system by creating two views of the same rotating
rlm@69 470 object from different angles and displaying both of those views in
rlm@69 471 JFrames.
rlm@69 472
rlm@69 473 You should see a rotating cube, and two windows,
rlm@69 474 each displaying a different view of the cube."
rlm@36 475 []
rlm@58 476 (let [candy
rlm@58 477 (box 1 1 1 :physical? false :color ColorRGBA/Blue)]
rlm@112 478 (world
rlm@112 479 (doto (Node.)
rlm@112 480 (.attachChild candy))
rlm@112 481 {}
rlm@112 482 (fn [world]
rlm@112 483 (let [cam (.clone (.getCamera world))
rlm@112 484 width (.getWidth cam)
rlm@112 485 height (.getHeight cam)]
rlm@169 486 (add-camera! world cam
rlm@215 487 (comp
rlm@215 488 (view-image
rlm@215 489 (File. "/home/r/proj/cortex/render/vision/1"))
rlm@215 490 BufferedImage!))
rlm@169 491 (add-camera! world
rlm@112 492 (doto (.clone cam)
rlm@112 493 (.setLocation (Vector3f. -10 0 0))
rlm@112 494 (.lookAt Vector3f/ZERO Vector3f/UNIT_Y))
rlm@215 495 (comp
rlm@215 496 (view-image
rlm@215 497 (File. "/home/r/proj/cortex/render/vision/2"))
rlm@215 498 BufferedImage!))
rlm@112 499 ;; This is here to restore the main view
rlm@112 500 ;; after the other views have completed processing
rlm@169 501 (add-camera! world (.getCamera world) no-op)))
rlm@112 502 (fn [world tpf]
rlm@112 503 (.rotate candy (* tpf 0.2) 0 0)))))
rlm@23 504 #+end_src
rlm@23 505
rlm@215 506 #+begin_html
rlm@215 507 <div class="figure">
rlm@215 508 <video controls="controls" width="755">
rlm@215 509 <source src="../video/spinning-cube.ogg" type="video/ogg"
rlm@215 510 preload="none" poster="../images/aurellem-1280x480.png" />
rlm@215 511 </video>
rlm@215 512 <p>A rotating cube viewed from two different perspectives.</p>
rlm@215 513 </div>
rlm@215 514 #+end_html
rlm@215 515
rlm@215 516 Creating multiple eyes like this can be used for stereoscopic vision
rlm@215 517 simulation in a single creature or for simulating multiple creatures,
rlm@215 518 each with their own sense of vision.
rlm@215 519
rlm@215 520 ** Adding Vision to the Worm
rlm@215 521
rlm@218 522 To the worm from the last post, I add a new node that describes its
rlm@215 523 eyes.
rlm@215 524
rlm@215 525 #+attr_html: width=755
rlm@215 526 #+caption: The worm with newly added empty nodes describing a single eye.
rlm@215 527 [[../images/worm-with-eye.png]]
rlm@215 528
rlm@215 529 The node highlighted in yellow is the root level "eyes" node. It has
rlm@218 530 a single child, highlighted in orange, which describes a single
rlm@218 531 eye. This is the "eye" node. It is placed so that the worm will have
rlm@218 532 an eye located in the center of the flat portion of its lower
rlm@218 533 hemispherical section.
rlm@218 534
rlm@218 535 The two nodes which are not highlighted describe the single joint of
rlm@218 536 the worm.
rlm@215 537
rlm@215 538 The metadata of the eye-node is:
rlm@215 539
rlm@215 540 #+begin_src clojure :results verbatim :exports both
rlm@215 541 (cortex.sense/meta-data
rlm@218 542 (.getChild (.getChild (cortex.test.body/worm) "eyes") "eye") "eye")
rlm@215 543 #+end_src
rlm@215 544
rlm@215 545 #+results:
rlm@215 546 : "(let [retina \"Models/test-creature/retina-small.png\"]
rlm@215 547 : {:all retina :red retina :green retina :blue retina})"
rlm@215 548
rlm@215 549 This is the approximation to the human eye described earlier.
rlm@215 550
rlm@216 551 #+name: test-2
rlm@215 552 #+begin_src clojure
rlm@215 553 (in-ns 'cortex.test.vision)
rlm@215 554
rlm@216 555 (defn change-color [obj color]
rlm@216 556 (println-repl obj)
rlm@216 557 (if obj
rlm@216 558 (.setColor (.getMaterial obj) "Color" color)))
rlm@216 559
rlm@216 560 (defn colored-cannon-ball [color]
rlm@216 561 (comp #(change-color % color)
rlm@216 562 (fire-cannon-ball)))
rlm@215 563
rlm@236 564 (defn test-worm-vision [record]
rlm@215 565 (let [the-worm (doto (worm)(body!))
rlm@215 566 vision (vision! the-worm)
rlm@215 567 vision-display (view-vision)
rlm@215 568 fix-display (gen-fix-display)
rlm@215 569 me (sphere 0.5 :color ColorRGBA/Blue :physical? false)
rlm@215 570 x-axis
rlm@215 571 (box 1 0.01 0.01 :physical? false :color ColorRGBA/Red
rlm@215 572 :position (Vector3f. 0 -5 0))
rlm@215 573 y-axis
rlm@215 574 (box 0.01 1 0.01 :physical? false :color ColorRGBA/Green
rlm@215 575 :position (Vector3f. 0 -5 0))
rlm@215 576 z-axis
rlm@215 577 (box 0.01 0.01 1 :physical? false :color ColorRGBA/Blue
rlm@216 578 :position (Vector3f. 0 -5 0))
rlm@216 579 timer (RatchetTimer. 60)]
rlm@215 580
rlm@215 581 (world (nodify [(floor) the-worm x-axis y-axis z-axis me])
rlm@216 582 (assoc standard-debug-controls
rlm@216 583 "key-r" (colored-cannon-ball ColorRGBA/Red)
rlm@216 584 "key-b" (colored-cannon-ball ColorRGBA/Blue)
rlm@216 585 "key-g" (colored-cannon-ball ColorRGBA/Green))
rlm@215 586 (fn [world]
rlm@215 587 (light-up-everything world)
rlm@216 588 (speed-up world)
rlm@216 589 (.setTimer world timer)
rlm@216 590 (display-dialated-time world timer)
rlm@215 591 ;; add a view from the worm's perspective
rlm@236 592 (if record
rlm@236 593 (Capture/captureVideo
rlm@236 594 world
rlm@236 595 (File.
rlm@236 596 "/home/r/proj/cortex/render/worm-vision/main-view")))
rlm@236 597
rlm@215 598 (add-camera!
rlm@215 599 world
rlm@215 600 (add-eye! the-worm
rlm@215 601 (.getChild
rlm@215 602 (.getChild the-worm "eyes") "eye"))
rlm@215 603 (comp
rlm@215 604 (view-image
rlm@236 605 (if record
rlm@236 606 (File.
rlm@236 607 "/home/r/proj/cortex/render/worm-vision/worm-view")))
rlm@215 608 BufferedImage!))
rlm@236 609
rlm@236 610 (set-gravity world Vector3f/ZERO))
rlm@216 611
rlm@215 612 (fn [world _ ]
rlm@215 613 (.setLocalTranslation me (.getLocation (.getCamera world)))
rlm@215 614 (vision-display
rlm@215 615 (map #(% world) vision)
rlm@236 616 (if record (File. "/home/r/proj/cortex/render/worm-vision")))
rlm@215 617 (fix-display world)))))
rlm@215 618 #+end_src
rlm@215 619
rlm@218 620 The world consists of the worm and a flat gray floor. I can shoot red,
rlm@218 621 green, blue and white cannonballs at the worm. The worm is initially
rlm@218 622 looking down at the floor, and there is no gravity. My perspective
rlm@218 623 (the Main View), the worm's perspective (Worm View) and the 4 sensor
rlm@218 624 channels that comprise the worm's eye are all saved frame-by-frame to
rlm@218 625 disk.
rlm@218 626
rlm@218 627 * Demonstration of Vision
rlm@218 628 #+begin_html
rlm@218 629 <div class="figure">
rlm@218 630 <video controls="controls" width="755">
rlm@218 631 <source src="../video/worm-vision.ogg" type="video/ogg"
rlm@218 632 preload="none" poster="../images/aurellem-1280x480.png" />
rlm@218 633 </video>
rlm@218 634 <p>Simulated Vision in a Virtual Environment</p>
rlm@218 635 </div>
rlm@218 636 #+end_html
rlm@218 637
rlm@218 638 ** Generate the Worm Video from Frames
rlm@216 639 #+name: magick2
rlm@216 640 #+begin_src clojure
rlm@216 641 (ns cortex.video.magick2
rlm@216 642 (:import java.io.File)
rlm@216 643 (:use clojure.contrib.shell-out))
rlm@216 644
rlm@216 645 (defn images [path]
rlm@216 646 (sort (rest (file-seq (File. path)))))
rlm@216 647
rlm@216 648 (def base "/home/r/proj/cortex/render/worm-vision/")
rlm@216 649
rlm@216 650 (defn pics [file]
rlm@216 651 (images (str base file)))
rlm@216 652
rlm@216 653 (defn combine-images []
rlm@216 654 (let [main-view (pics "main-view")
rlm@216 655 worm-view (pics "worm-view")
rlm@216 656 blue (pics "0")
rlm@216 657 green (pics "1")
rlm@216 658 red (pics "2")
rlm@216 659 gray (pics "3")
rlm@216 660 blender (let [b-pics (pics "blender")]
rlm@216 661 (concat b-pics (repeat 9001 (last b-pics))))
rlm@216 662 background (repeat 9001 (File. (str base "background.png")))
rlm@216 663 targets (map
rlm@216 664 #(File. (str base "out/" (format "%07d.png" %)))
rlm@216 665 (range 0 (count main-view)))]
rlm@216 666 (dorun
rlm@216 667 (pmap
rlm@216 668 (comp
rlm@216 669 (fn [[background main-view worm-view red green blue gray blender target]]
rlm@216 670 (println target)
rlm@216 671 (sh "convert"
rlm@216 672 background
rlm@216 673 main-view "-geometry" "+18+17" "-composite"
rlm@216 674 worm-view "-geometry" "+677+17" "-composite"
rlm@216 675 green "-geometry" "+685+430" "-composite"
rlm@216 676 red "-geometry" "+788+430" "-composite"
rlm@216 677 blue "-geometry" "+894+430" "-composite"
rlm@216 678 gray "-geometry" "+1000+430" "-composite"
rlm@216 679 blender "-geometry" "+0+0" "-composite"
rlm@216 680 target))
rlm@216 681 (fn [& args] (map #(.getCanonicalPath %) args)))
rlm@216 682 background main-view worm-view red green blue gray blender targets))))
rlm@216 683 #+end_src
rlm@216 684
rlm@216 685 #+begin_src sh :results silent
rlm@216 686 cd /home/r/proj/cortex/render/worm-vision
rlm@216 687 ffmpeg -r 25 -b 9001k -i out/%07d.png -vcodec libtheora worm-vision.ogg
rlm@216 688 #+end_src
rlm@236 689
rlm@215 690 * Headers
rlm@215 691
rlm@213 692 #+name: vision-header
rlm@213 693 #+begin_src clojure
rlm@213 694 (ns cortex.vision
rlm@213 695 "Simulate the sense of vision in jMonkeyEngine3. Enables multiple
rlm@213 696 eyes from different positions to observe the same world, and pass
rlm@213 697 the observed data to any arbitray function. Automatically reads
rlm@216 698 eye-nodes from specially prepared blender files and instantiates
rlm@213 699 them in the world as actual eyes."
rlm@213 700 {:author "Robert McIntyre"}
rlm@213 701 (:use (cortex world sense util))
rlm@213 702 (:use clojure.contrib.def)
rlm@213 703 (:import com.jme3.post.SceneProcessor)
rlm@237 704 (:import (com.jme3.util BufferUtils Screenshots))
rlm@213 705 (:import java.nio.ByteBuffer)
rlm@213 706 (:import java.awt.image.BufferedImage)
rlm@213 707 (:import (com.jme3.renderer ViewPort Camera))
rlm@216 708 (:import (com.jme3.math ColorRGBA Vector3f Matrix3f))
rlm@213 709 (:import com.jme3.renderer.Renderer)
rlm@213 710 (:import com.jme3.app.Application)
rlm@213 711 (:import com.jme3.texture.FrameBuffer)
rlm@213 712 (:import (com.jme3.scene Node Spatial)))
rlm@213 713 #+end_src
rlm@112 714
rlm@215 715 #+name: test-header
rlm@215 716 #+begin_src clojure
rlm@215 717 (ns cortex.test.vision
rlm@215 718 (:use (cortex world sense util body vision))
rlm@215 719 (:use cortex.test.body)
rlm@215 720 (:import java.awt.image.BufferedImage)
rlm@215 721 (:import javax.swing.JPanel)
rlm@215 722 (:import javax.swing.SwingUtilities)
rlm@215 723 (:import java.awt.Dimension)
rlm@215 724 (:import javax.swing.JFrame)
rlm@215 725 (:import com.jme3.math.ColorRGBA)
rlm@215 726 (:import com.jme3.scene.Node)
rlm@215 727 (:import com.jme3.math.Vector3f)
rlm@216 728 (:import java.io.File)
rlm@216 729 (:import (com.aurellem.capture Capture RatchetTimer)))
rlm@215 730 #+end_src
rlm@215 731
rlm@216 732 * Onward!
rlm@216 733 - As a neat bonus, this idea behind simulated vision also enables one
rlm@216 734 to [[../../cortex/html/capture-video.html][capture live video feeds from jMonkeyEngine]].
rlm@216 735 - Now that we have vision, it's time to tackle [[./hearing.org][hearing]].
rlm@215 736
rlm@216 737 * Source Listing
rlm@216 738 - [[../src/cortex/vision.clj][cortex.vision]]
rlm@216 739 - [[../src/cortex/test/vision.clj][cortex.test.vision]]
rlm@216 740 - [[../src/cortex/video/magick2.clj][cortex.video.magick2]]
rlm@216 741 - [[../assets/Models/subtitles/worm-vision-subtitles.blend][worm-vision-subtitles.blend]]
rlm@216 742 #+html: <ul> <li> <a href="../org/sense.org">This org file</a> </li> </ul>
rlm@216 743 - [[http://hg.bortreb.com ][source-repository]]
rlm@216 744
rlm@35 745
rlm@24 746
rlm@212 747 * COMMENT Generate Source
rlm@34 748 #+begin_src clojure :tangle ../src/cortex/vision.clj
rlm@216 749 <<vision-header>>
rlm@216 750 <<pipeline-1>>
rlm@216 751 <<pipeline-2>>
rlm@216 752 <<retina>>
rlm@216 753 <<add-eye>>
rlm@216 754 <<sensitivity>>
rlm@216 755 <<eye-node>>
rlm@216 756 <<add-camera>>
rlm@216 757 <<kernel>>
rlm@216 758 <<main>>
rlm@216 759 <<display>>
rlm@24 760 #+end_src
rlm@24 761
rlm@68 762 #+begin_src clojure :tangle ../src/cortex/test/vision.clj
rlm@215 763 <<test-header>>
rlm@215 764 <<test-1>>
rlm@216 765 <<test-2>>
rlm@24 766 #+end_src
rlm@216 767
rlm@216 768 #+begin_src clojure :tangle ../src/cortex/video/magick2.clj
rlm@216 769 <<magick2>>
rlm@216 770 #+end_src