added docs, decoding colour sort-of works

chanokin · Feb 13, 2020 · 3831649 · 3831649
1 parent 0ce1fbf
commit 3831649
Show file tree

Hide file tree

Showing 9 changed files with 238 additions and 364 deletions.
diff --git a/README.md b/README.md
@@ -3,7 +3,7 @@
 
 **Usage:**
 
-   `python convert.py dataset path/to/input/files [options]`
+    python convert.py dataset path/to/input/files [options]
 
     --timestep         Timestep which will be used in the simulations. How many spikes will be emmited 
                        at each timestep can be set with --spikes_per_bin
@@ -22,8 +22,11 @@
 1. ___Databases___ in this repository are the property of their authors:
 
    * __[CIFAR-10](https://www.cs.toronto.edu/~kriz/cifar.html)__: [Alex Krizhevsky, (2009). Learning Multiple Layers of Features from Tiny Images](https://www.cs.toronto.edu/~kriz/learning-features-2009-TR.pdf)
+      * This database has color images which get transformed into a YUV encoding. The conversion outpus rank-order coded spikes for each of the images (grayscale [Y], blue-yellow [U],  red-green [V])
    * __[MNIST](http://yann.lecun.com/exdb/mnist/)__: [Y. LeCun, L. Bottou, Y. Bengio, and P. Haffner. Gradient-based learning applied to document recognition.](http://yann.lecun.com/exdb/publis/index.html#lecun-98) Proceedings of the IEEE, 86(11):2278-2324, November 1998.
+      * The images in the dataset are converted so that background value is 0 and digit regions values are 255. The output is rank-order coded spikes from the 28x28 images.
    * __[OMNIGLOT](https://github.com/brendenlake/omniglot)__: [Lake, B. M., Salakhutdinov, R., and Tenenbaum, J. B. (2015). Human-level concept learning through probabilistic program induction.](http://www.sciencemag.org/content/350/6266/1332.short) _Science_, 350(6266), 1332-1338.
+      * The images in the dataset are converted so that background value is 0 and digit regions values are greater than 0; furthermore, the images are scaled using the `--scaling` parameter. The output is rank-order coded spikes from the scaled images.
 
 2. The (grayscale) ___transformation method___ was:
    * Created by __[Basabdatta Sen Bhattacharya](https://sites.google.com/site/bsenbhattacharya/)__

diff --git a/cifar_convert.py b/cifar_convert.py
@@ -81,6 +81,7 @@ def cifar_convert(data, out_dir, timestep, spikes_per_bin=1, skip_existing=True)
             kernels=FOCAL_S.kernels.full_kernels,
             bmg_image=bmg, bmg_spikes=bmg_spikes, bmg_spk_src=bmg_spk_src,
             rmg_image=rmg, rmg_spikes=rmg_spikes, rmg_spk_src=rmg_spk_src,
+            wred=WRED, wgreen=WGREEN, wblue=WBLUE,
             )
 
     print("\tDone with batch!\n")

diff --git a/color_encoding_notebooks/color_encoding_yuv.ipynb b/color_encoding_notebooks/color_encoding_yuv.ipynb
@@ -1,7 +1,7 @@
 {
  "metadata": {
   "name": "",
-  "signature": "sha256:bd7595ac957ae27bcfbde87f0b47068141d34b30e8a8aafb55c7a9a18d43eca5"
+  "signature": "sha256:033d10c0cc4f1d0aefd3313964b60ed7fc2f64373f118ab69e97abdfabe094c8"
  },
  "nbformat": 3,
  "nbformat_minor": 0,
@@ -17,7 +17,7 @@
       "import matplotlib.pyplot as plt\n",
       "import os\n",
       "import glob\n",
-      "from focal import Focal, focal_to_spike, spike_trains_to_images_g\n",
+      "# from focal import Focal, focal_to_spike, spike_trains_to_images_g\n",
       "from scipy.signal import convolve2d\n",
       "from scipy import misc\n",
       "import cv2\n",
@@ -88,7 +88,7 @@
      "cell_type": "code",
      "collapsed": false,
      "input": [
-      "image_files = sorted( glob.glob('./test_images/*.png') )\n",
+      "image_files = sorted( glob.glob('../test_images/*.png') )\n",
       "face = cv2.cvtColor( cv2.imread(image_files[7]), cv2.COLOR_BGR2RGB ).astype('float')\n",
       "\n",
       "face = misc.face() / 255.0\n",
@@ -230,7 +230,7 @@
        ]
       }
      ],
-     "prompt_number": 7
+     "prompt_number": 6
     },
     {
      "cell_type": "code",
@@ -239,7 +239,7 @@
      "language": "python",
      "metadata": {},
      "outputs": [],
-     "prompt_number": 7
+     "prompt_number": 6
     },
     {
      "cell_type": "code",
@@ -258,7 +258,7 @@
      "language": "python",
      "metadata": {},
      "outputs": [],
-     "prompt_number": 8
+     "prompt_number": 7
     },
     {
      "cell_type": "code",
@@ -298,7 +298,7 @@
        ]
       }
      ],
-     "prompt_number": 23
+     "prompt_number": 8
     },
     {
      "cell_type": "code",
@@ -321,7 +321,7 @@
        ]
       }
      ],
-     "prompt_number": 24
+     "prompt_number": 9
     },
     {
      "cell_type": "code",
@@ -420,7 +420,7 @@
        ]
       }
      ],
-     "prompt_number": 45
+     "prompt_number": 10
     },
     {
      "cell_type": "code",
@@ -439,7 +439,7 @@
      "language": "python",
      "metadata": {},
      "outputs": [],
-     "prompt_number": 46
+     "prompt_number": 11
     },
     {
      "cell_type": "code",
@@ -499,7 +499,7 @@
        ]
       }
      ],
-     "prompt_number": 47
+     "prompt_number": 14
     },
     {
      "cell_type": "code",

diff --git a/convert.py b/convert.py
@@ -20,7 +20,7 @@
 parser.add_argument('--spikes_per_bin', type=int, default=1,
                     help='How many spikes per timestep will be emmited. Note that more '
                     ' than one is not standard rank-order encoding.')
-parser.add_argument('--scaling', type=float, default=1.0,
+parser.add_argument('--scaling', type=float, default=0.54,
                     help='Scaling applied to the input image (only supported by the '
                     ' Omniglot dataset)')
 args = parser.parse_args()

diff --git a/focal/convolution.py b/focal/convolution.py
@@ -136,13 +136,13 @@ def get_subsample_keepers(self, cell_type):
         if cell_type == 3:
             # col_keep = 7
             # row_keep = 7
-            col_keep = 5
-            row_keep = 5
+            col_keep = 3
+            row_keep = 3
         elif cell_type == 2:
             # col_keep = 5
             # row_keep = 3
-            col_keep = 4
-            row_keep = 4
+            col_keep = 3
+            row_keep = 3
         elif cell_type == 1:
             col_keep = 1
             row_keep = 1

diff --git a/load_and_read_notebooks/test_read_cifar10.ipynb b/load_and_read_notebooks/test_read_cifar10.ipynb
diff --git a/load_and_read_notebooks/test_read_mnist.ipynb b/load_and_read_notebooks/test_read_mnist.ipynb
diff --git a/load_and_read_notebooks/test_read_omniglot.ipynb b/load_and_read_notebooks/test_read_omniglot.ipynb
diff --git a/omniglot_convert.py b/omniglot_convert.py
@@ -63,7 +63,10 @@ def omniglot_convert(file_dict, out_dir, timestep, spikes_per_bin=1, skip_existi
 
     h, w = file_dict[alpha][char][0].shape
     if scaling != 1.0:
+        sys.stdout.write('\t\tscaling input image from shape {}'.format((h, w)))
         h, w = int(h * scaling), int(w * scaling)
+        sys.stdout.write(' to {}\n\n'.format((h, w)))
+        sys.stdout.flush()
 
     s_img = np.zeros((h, w))
     n_processed = 0