|
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@23
|
11
|
|
rlm@34
|
12 I want to make creatures with eyes. Each eye can be independely moved
|
|
rlm@34
|
13 and should see its own version of the world depending on where it is.
|
|
rlm@151
|
14
|
|
rlm@151
|
15 Here's how vision will work.
|
|
rlm@151
|
16
|
|
rlm@151
|
17 Make the continuation in scene-processor take FrameBuffer,
|
|
rlm@151
|
18 byte-buffer, BufferedImage already sized to the correct
|
|
rlm@151
|
19 dimensions. the continuation will decide wether to "mix" them
|
|
rlm@151
|
20 into the BufferedImage, lazily ignore them, or mix them halfway
|
|
rlm@151
|
21 and call c/graphics card routines.
|
|
rlm@151
|
22
|
|
rlm@151
|
23 (vision creature) will take an optional :skip argument which will
|
|
rlm@151
|
24 inform the continuations in scene processor to skip the given
|
|
rlm@151
|
25 number of cycles 0 means that no cycles will be skipped.
|
|
rlm@151
|
26
|
|
rlm@151
|
27 (vision creature) will return [init-functions sensor-functions].
|
|
rlm@151
|
28 The init-functions are each single-arg functions that take the
|
|
rlm@151
|
29 world and register the cameras and must each be called before the
|
|
rlm@151
|
30 corresponding sensor-functions. Each init-function returns the
|
|
rlm@151
|
31 viewport for that eye which can be manipulated, saved, etc. Each
|
|
rlm@151
|
32 sensor-function is a thunk and will return data in the same
|
|
rlm@151
|
33 format as the tactile-sensor functions the structure is
|
|
rlm@151
|
34 [topology, sensor-data]. Internally, these sensor-functions
|
|
rlm@151
|
35 maintain a reference to sensor-data which is periodically updated
|
|
rlm@151
|
36 by the continuation function established by its init-function.
|
|
rlm@151
|
37 They can be queried every cycle, but their information may not
|
|
rlm@151
|
38 necessairly be different every cycle.
|
|
rlm@151
|
39
|
|
rlm@151
|
40 Each eye in the creature in blender will work the same way as
|
|
rlm@151
|
41 joints -- a zero dimensional object with no geometry whose local
|
|
rlm@151
|
42 coordinate system determines the orientation of the resulting
|
|
rlm@151
|
43 eye. All eyes will have a parent named "eyes" just as all joints
|
|
rlm@151
|
44 have a parent named "joints". The resulting camera will be a
|
|
rlm@151
|
45 ChaseCamera or a CameraNode bound to the geo that is closest to
|
|
rlm@151
|
46 the eye marker. The eye marker will contain the metadata for the
|
|
rlm@151
|
47 eye, and will be moved by it's bound geometry. The dimensions of
|
|
rlm@151
|
48 the eye's camera are equal to the dimensions of the eye's "UV"
|
|
rlm@151
|
49 map.
|
|
rlm@151
|
50
|
|
rlm@66
|
51 #+name: eyes
|
|
rlm@23
|
52 #+begin_src clojure
|
|
rlm@34
|
53 (ns cortex.vision
|
|
rlm@34
|
54 "Simulate the sense of vision in jMonkeyEngine3. Enables multiple
|
|
rlm@34
|
55 eyes from different positions to observe the same world, and pass
|
|
rlm@172
|
56 the observed data to any arbitray function. Automatically reads
|
|
rlm@172
|
57 eye-nodes from specially prepared blender files and instanttiates
|
|
rlm@172
|
58 them in the world as actual eyes."
|
|
rlm@34
|
59 {:author "Robert McIntyre"}
|
|
rlm@151
|
60 (:use (cortex world sense util))
|
|
rlm@167
|
61 (:use clojure.contrib.def)
|
|
rlm@34
|
62 (:import com.jme3.post.SceneProcessor)
|
|
rlm@113
|
63 (:import (com.jme3.util BufferUtils Screenshots))
|
|
rlm@34
|
64 (:import java.nio.ByteBuffer)
|
|
rlm@34
|
65 (:import java.awt.image.BufferedImage)
|
|
rlm@172
|
66 (:import (com.jme3.renderer ViewPort Camera))
|
|
rlm@113
|
67 (:import com.jme3.math.ColorRGBA)
|
|
rlm@151
|
68 (:import com.jme3.renderer.Renderer)
|
|
rlm@172
|
69 (:import com.jme3.app.Application)
|
|
rlm@172
|
70 (:import com.jme3.texture.FrameBuffer)
|
|
rlm@172
|
71 (:import (com.jme3.scene Node Spatial)))
|
|
rlm@113
|
72
|
|
rlm@113
|
73 (defn vision-pipeline
|
|
rlm@34
|
74 "Create a SceneProcessor object which wraps a vision processing
|
|
rlm@113
|
75 continuation function. The continuation is a function that takes
|
|
rlm@113
|
76 [#^Renderer r #^FrameBuffer fb #^ByteBuffer b #^BufferedImage bi],
|
|
rlm@113
|
77 each of which has already been appropiately sized."
|
|
rlm@23
|
78 [continuation]
|
|
rlm@23
|
79 (let [byte-buffer (atom nil)
|
|
rlm@113
|
80 renderer (atom nil)
|
|
rlm@113
|
81 image (atom nil)]
|
|
rlm@23
|
82 (proxy [SceneProcessor] []
|
|
rlm@23
|
83 (initialize
|
|
rlm@23
|
84 [renderManager viewPort]
|
|
rlm@23
|
85 (let [cam (.getCamera viewPort)
|
|
rlm@23
|
86 width (.getWidth cam)
|
|
rlm@23
|
87 height (.getHeight cam)]
|
|
rlm@23
|
88 (reset! renderer (.getRenderer renderManager))
|
|
rlm@23
|
89 (reset! byte-buffer
|
|
rlm@23
|
90 (BufferUtils/createByteBuffer
|
|
rlm@113
|
91 (* width height 4)))
|
|
rlm@113
|
92 (reset! image (BufferedImage.
|
|
rlm@113
|
93 width height
|
|
rlm@113
|
94 BufferedImage/TYPE_4BYTE_ABGR))))
|
|
rlm@23
|
95 (isInitialized [] (not (nil? @byte-buffer)))
|
|
rlm@23
|
96 (reshape [_ _ _])
|
|
rlm@23
|
97 (preFrame [_])
|
|
rlm@23
|
98 (postQueue [_])
|
|
rlm@23
|
99 (postFrame
|
|
rlm@23
|
100 [#^FrameBuffer fb]
|
|
rlm@23
|
101 (.clear @byte-buffer)
|
|
rlm@113
|
102 (continuation @renderer fb @byte-buffer @image))
|
|
rlm@23
|
103 (cleanup []))))
|
|
rlm@23
|
104
|
|
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@112
|
123
|
|
rlm@169
|
124 (defn add-camera!
|
|
rlm@169
|
125 "Add a camera to the world, calling continuation on every frame
|
|
rlm@34
|
126 produced."
|
|
rlm@167
|
127 [#^Application world camera continuation]
|
|
rlm@23
|
128 (let [width (.getWidth camera)
|
|
rlm@23
|
129 height (.getHeight camera)
|
|
rlm@23
|
130 render-manager (.getRenderManager world)
|
|
rlm@23
|
131 viewport (.createMainView render-manager "eye-view" camera)]
|
|
rlm@23
|
132 (doto viewport
|
|
rlm@23
|
133 (.setClearFlags true true true)
|
|
rlm@112
|
134 (.setBackgroundColor ColorRGBA/Black)
|
|
rlm@113
|
135 (.addProcessor (vision-pipeline continuation))
|
|
rlm@23
|
136 (.attachScene (.getRootNode world)))))
|
|
rlm@151
|
137
|
|
rlm@169
|
138 (defn retina-sensor-profile
|
|
rlm@151
|
139 "Return a map of pixel selection functions to BufferedImages
|
|
rlm@169
|
140 describing the distribution of light-sensitive components of this
|
|
rlm@169
|
141 eye. Each function creates an integer from the rgb values found in
|
|
rlm@169
|
142 the pixel. :red, :green, :blue, :gray are already defined as
|
|
rlm@169
|
143 extracting the red, green, blue, and average components
|
|
rlm@151
|
144 respectively."
|
|
rlm@151
|
145 [#^Spatial eye]
|
|
rlm@151
|
146 (if-let [eye-map (meta-data eye "eye")]
|
|
rlm@151
|
147 (map-vals
|
|
rlm@167
|
148 load-image
|
|
rlm@151
|
149 (eval (read-string eye-map)))))
|
|
rlm@151
|
150
|
|
rlm@151
|
151 (defn eye-dimensions
|
|
rlm@169
|
152 "Returns [width, height] specified in the metadata of the eye"
|
|
rlm@151
|
153 [#^Spatial eye]
|
|
rlm@151
|
154 (let [dimensions
|
|
rlm@151
|
155 (map #(vector (.getWidth %) (.getHeight %))
|
|
rlm@169
|
156 (vals (retina-sensor-profile eye)))]
|
|
rlm@151
|
157 [(apply max (map first dimensions))
|
|
rlm@151
|
158 (apply max (map second dimensions))]))
|
|
rlm@151
|
159
|
|
rlm@167
|
160 (defvar
|
|
rlm@167
|
161 ^{:arglists '([creature])}
|
|
rlm@167
|
162 eyes
|
|
rlm@167
|
163 (sense-nodes "eyes")
|
|
rlm@167
|
164 "Return the children of the creature's \"eyes\" node.")
|
|
rlm@151
|
165
|
|
rlm@169
|
166 (defn add-eye!
|
|
rlm@169
|
167 "Create a Camera centered on the current position of 'eye which
|
|
rlm@169
|
168 follows the closest physical node in 'creature and sends visual
|
|
rlm@169
|
169 data to 'continuation."
|
|
rlm@151
|
170 [#^Node creature #^Spatial eye]
|
|
rlm@151
|
171 (let [target (closest-node creature eye)
|
|
rlm@151
|
172 [cam-width cam-height] (eye-dimensions eye)
|
|
rlm@151
|
173 cam (Camera. cam-width cam-height)]
|
|
rlm@151
|
174 (.setLocation cam (.getWorldTranslation eye))
|
|
rlm@151
|
175 (.setRotation cam (.getWorldRotation eye))
|
|
rlm@151
|
176 (.setFrustumPerspective
|
|
rlm@151
|
177 cam 45 (/ (.getWidth cam) (.getHeight cam))
|
|
rlm@151
|
178 1 1000)
|
|
rlm@151
|
179 (bind-sense target cam)
|
|
rlm@151
|
180 cam))
|
|
rlm@151
|
181
|
|
rlm@172
|
182 (defvar color-channel-presets
|
|
rlm@151
|
183 {:all 0xFFFFFF
|
|
rlm@151
|
184 :red 0xFF0000
|
|
rlm@151
|
185 :blue 0x0000FF
|
|
rlm@172
|
186 :green 0x00FF00}
|
|
rlm@172
|
187 "Bitmasks for common RGB color channels")
|
|
rlm@151
|
188
|
|
rlm@169
|
189 (defn vision-fn
|
|
rlm@171
|
190 "Returns a list of functions, each of which will return a color
|
|
rlm@171
|
191 channel's worth of visual information when called inside a running
|
|
rlm@171
|
192 simulation."
|
|
rlm@151
|
193 [#^Node creature #^Spatial eye & {skip :skip :or {skip 0}}]
|
|
rlm@169
|
194 (let [retinal-map (retina-sensor-profile eye)
|
|
rlm@169
|
195 camera (add-eye! creature eye)
|
|
rlm@151
|
196 vision-image
|
|
rlm@151
|
197 (atom
|
|
rlm@151
|
198 (BufferedImage. (.getWidth camera)
|
|
rlm@151
|
199 (.getHeight camera)
|
|
rlm@170
|
200 BufferedImage/TYPE_BYTE_BINARY))
|
|
rlm@170
|
201 register-eye!
|
|
rlm@170
|
202 (runonce
|
|
rlm@170
|
203 (fn [world]
|
|
rlm@170
|
204 (add-camera!
|
|
rlm@170
|
205 world camera
|
|
rlm@170
|
206 (let [counter (atom 0)]
|
|
rlm@170
|
207 (fn [r fb bb bi]
|
|
rlm@170
|
208 (if (zero? (rem (swap! counter inc) (inc skip)))
|
|
rlm@170
|
209 (reset! vision-image
|
|
rlm@170
|
210 (BufferedImage! r fb bb bi))))))))]
|
|
rlm@151
|
211 (vec
|
|
rlm@151
|
212 (map
|
|
rlm@151
|
213 (fn [[key image]]
|
|
rlm@151
|
214 (let [whites (white-coordinates image)
|
|
rlm@151
|
215 topology (vec (collapse whites))
|
|
rlm@172
|
216 mask (color-channel-presets key)]
|
|
rlm@170
|
217 (fn [world]
|
|
rlm@170
|
218 (register-eye! world)
|
|
rlm@151
|
219 (vector
|
|
rlm@151
|
220 topology
|
|
rlm@151
|
221 (vec
|
|
rlm@151
|
222 (for [[x y] whites]
|
|
rlm@151
|
223 (bit-and
|
|
rlm@151
|
224 mask (.getRGB @vision-image x y))))))))
|
|
rlm@170
|
225 retinal-map))))
|
|
rlm@151
|
226
|
|
rlm@170
|
227
|
|
rlm@170
|
228 ;; TODO maybe should add a viewport-manipulation function to
|
|
rlm@170
|
229 ;; automatically change viewport settings, attach shadow filters, etc.
|
|
rlm@170
|
230
|
|
rlm@170
|
231 (defn vision!
|
|
rlm@170
|
232 "Returns a function which returns visual sensory data when called
|
|
rlm@170
|
233 inside a running simulation"
|
|
rlm@151
|
234 [#^Node creature & {skip :skip :or {skip 0}}]
|
|
rlm@151
|
235 (reduce
|
|
rlm@170
|
236 concat
|
|
rlm@167
|
237 (for [eye (eyes creature)]
|
|
rlm@169
|
238 (vision-fn creature eye))))
|
|
rlm@151
|
239
|
|
rlm@189
|
240 (defn view-vision
|
|
rlm@189
|
241 "Creates a function which accepts a list of visual sensor-data and
|
|
rlm@189
|
242 displays each element of the list to the screen."
|
|
rlm@189
|
243 []
|
|
rlm@188
|
244 (view-sense
|
|
rlm@188
|
245 (fn
|
|
rlm@188
|
246 [[coords sensor-data]]
|
|
rlm@188
|
247 (let [image (points->image coords)]
|
|
rlm@188
|
248 (dorun
|
|
rlm@188
|
249 (for [i (range (count coords))]
|
|
rlm@188
|
250 (.setRGB image ((coords i) 0) ((coords i) 1)
|
|
rlm@188
|
251 (sensor-data i))))
|
|
rlm@189
|
252 image))))
|
|
rlm@188
|
253
|
|
rlm@34
|
254 #+end_src
|
|
rlm@23
|
255
|
|
rlm@112
|
256
|
|
rlm@34
|
257 Note the use of continuation passing style for connecting the eye to a
|
|
rlm@34
|
258 function to process the output. You can create any number of eyes, and
|
|
rlm@34
|
259 each of them will see the world from their own =Camera=. Once every
|
|
rlm@34
|
260 frame, the rendered image is copied to a =BufferedImage=, and that
|
|
rlm@34
|
261 data is sent off to the continuation function. Moving the =Camera=
|
|
rlm@34
|
262 which was used to create the eye will change what the eye sees.
|
|
rlm@23
|
263
|
|
rlm@34
|
264 * Example
|
|
rlm@23
|
265
|
|
rlm@66
|
266 #+name: test-vision
|
|
rlm@23
|
267 #+begin_src clojure
|
|
rlm@68
|
268 (ns cortex.test.vision
|
|
rlm@34
|
269 (:use (cortex world util vision))
|
|
rlm@34
|
270 (:import java.awt.image.BufferedImage)
|
|
rlm@34
|
271 (:import javax.swing.JPanel)
|
|
rlm@34
|
272 (:import javax.swing.SwingUtilities)
|
|
rlm@34
|
273 (:import java.awt.Dimension)
|
|
rlm@34
|
274 (:import javax.swing.JFrame)
|
|
rlm@34
|
275 (:import com.jme3.math.ColorRGBA)
|
|
rlm@45
|
276 (:import com.jme3.scene.Node)
|
|
rlm@113
|
277 (:import com.jme3.math.Vector3f))
|
|
rlm@23
|
278
|
|
rlm@36
|
279 (defn test-two-eyes
|
|
rlm@69
|
280 "Testing vision:
|
|
rlm@69
|
281 Tests the vision system by creating two views of the same rotating
|
|
rlm@69
|
282 object from different angles and displaying both of those views in
|
|
rlm@69
|
283 JFrames.
|
|
rlm@69
|
284
|
|
rlm@69
|
285 You should see a rotating cube, and two windows,
|
|
rlm@69
|
286 each displaying a different view of the cube."
|
|
rlm@36
|
287 []
|
|
rlm@58
|
288 (let [candy
|
|
rlm@58
|
289 (box 1 1 1 :physical? false :color ColorRGBA/Blue)]
|
|
rlm@112
|
290 (world
|
|
rlm@112
|
291 (doto (Node.)
|
|
rlm@112
|
292 (.attachChild candy))
|
|
rlm@112
|
293 {}
|
|
rlm@112
|
294 (fn [world]
|
|
rlm@112
|
295 (let [cam (.clone (.getCamera world))
|
|
rlm@112
|
296 width (.getWidth cam)
|
|
rlm@112
|
297 height (.getHeight cam)]
|
|
rlm@169
|
298 (add-camera! world cam
|
|
rlm@113
|
299 ;;no-op
|
|
rlm@113
|
300 (comp (view-image) BufferedImage!)
|
|
rlm@112
|
301 )
|
|
rlm@169
|
302 (add-camera! world
|
|
rlm@112
|
303 (doto (.clone cam)
|
|
rlm@112
|
304 (.setLocation (Vector3f. -10 0 0))
|
|
rlm@112
|
305 (.lookAt Vector3f/ZERO Vector3f/UNIT_Y))
|
|
rlm@113
|
306 ;;no-op
|
|
rlm@113
|
307 (comp (view-image) BufferedImage!))
|
|
rlm@112
|
308 ;; This is here to restore the main view
|
|
rlm@112
|
309 ;; after the other views have completed processing
|
|
rlm@169
|
310 (add-camera! world (.getCamera world) no-op)))
|
|
rlm@112
|
311 (fn [world tpf]
|
|
rlm@112
|
312 (.rotate candy (* tpf 0.2) 0 0)))))
|
|
rlm@23
|
313 #+end_src
|
|
rlm@23
|
314
|
|
rlm@112
|
315 #+results: test-vision
|
|
rlm@112
|
316 : #'cortex.test.vision/test-two-eyes
|
|
rlm@112
|
317
|
|
rlm@34
|
318 The example code will create two videos of the same rotating object
|
|
rlm@34
|
319 from different angles. It can be used both for stereoscopic vision
|
|
rlm@34
|
320 simulation or for simulating multiple creatures, each with their own
|
|
rlm@34
|
321 sense of vision.
|
|
rlm@24
|
322
|
|
rlm@35
|
323 - As a neat bonus, this idea behind simulated vision also enables one
|
|
rlm@35
|
324 to [[../../cortex/html/capture-video.html][capture live video feeds from jMonkeyEngine]].
|
|
rlm@35
|
325
|
|
rlm@24
|
326
|
|
rlm@24
|
327 * COMMENT code generation
|
|
rlm@34
|
328 #+begin_src clojure :tangle ../src/cortex/vision.clj
|
|
rlm@24
|
329 <<eyes>>
|
|
rlm@24
|
330 #+end_src
|
|
rlm@24
|
331
|
|
rlm@68
|
332 #+begin_src clojure :tangle ../src/cortex/test/vision.clj
|
|
rlm@24
|
333 <<test-vision>>
|
|
rlm@24
|
334 #+end_src
|
