Improve PNG compression in add_image_summary (#995)

OpenCV does not fully compress PNGs by default. This additional parameter tells OpenCV to use the maximum PNG compression level for encoding images. This should reduce the size of saved Tensorboard images particularly for highly compressible domains such as segmentations.

tensorpack/tfutils/summary.py

@@ -68,6 +68,7 @@ def create_image_summary(name, val):
     n, h, w, c = val.shape
     val = val.astype('uint8')
     s = tf.Summary()
+    imparams = [cv2.IMWRITE_PNG_COMPRESSION, 9]
     for k in range(n):
         arr = val[k]
         # CV2 will only write correctly in BGR chanel order
@@ -76,8 +77,7 @@ def create_image_summary(name, val):
         elif c == 4:
             arr = cv2.cvtColor(arr, cv2.COLOR_RGBA2BGRA)
         tag = name if n == 1 else '{}/{}'.format(name, k)
-
-        retval, img_str = cv2.imencode('.png', arr)
+        retval, img_str = cv2.imencode('.png', arr, imparams)
         if not retval:
             # Encoding has failed.
             continue