gif added, README updated

.gitignore

@@ -1,4 +1,5 @@
 Train/
 *.h5
 logs/
-*.zip
+*.zip
+*.mp4

README.md

@@ -13,6 +13,17 @@ The [lyft Perception challenge](https://www.udacity.com/lyft-challenge) in assoc
 ## Approach
 Although it was a segmentation problem and did not require instance segmentation, I went ahead with [MASK-RCNN](https://arxiv.org/pdf/1703.06870.pdf) as it was the state of the art algorithm in image segmentation and I was always intrigued to learn about it. Also I started on *28th*, just after finishing my first term, so transfer learning was my only shot. :sweat:
 
+
+*Click to Watch on youtube*
+
+
+<div style="text-align:center;">
+	<a href=https://www.youtube.com/watch?v=Q56fzNjmYKc>
+		<img src=./assets/final.gif width="640" height="480" >
+	</a>
+</div>
+
+
 #### Mask-RCNN (A brief overview)
 
 Mask-RCNN, also known as [Detectron](https://github.com/facebookresearch/Detectron) is a research platform for object detection developed by facebookresearch. It is mainly a modification of Faster RCNN with a segmentation branch parallel to class predictor and bounding box regressor. The vanilla ResNet is used in an FPN setting as a backbone to Faster RCNN so that features can be extracted at multiple levels of the feature pyramid
@@ -66,7 +77,7 @@ def process_labels(self,labels):
 ```
 
 #### MaskRCNN Configuration
-For this application Resnet-50 was used by setting `BACKBONE = "resnet50"` in config.
+For this application Resnet-50 was used by setting `BACKBONE = "resnet50"` in config. `NUM_CLASSES = 1+2` for 2 classes (car,road) and `IMAGE_MAX_DIM = 1024` as image dimensions are 800x600.
 
 ```python
 class LyftChallengeConfig(Config):

inference.py

@@ -27,6 +27,7 @@
 import mrcnn.model as modellib
 from mrcnn import visualize
 from mrcnn.model import log
+from moviepy.editor import VideoFileClip
 
 
 
@@ -165,21 +166,11 @@ def segment_images_batch(original_images):
     except KeyboardInterrupt as e:
         break
 
-# video_len = video.shape[0]
-# offset=1
-# a = time.time()
-# for idx in range(0,video_len,4):
-#     # if video_len-idx >4:
-#     #     offset = 4
-#     # else: 
-#     #     offset = video_len-idx
-#     # model.config.IMAGES_PER_GPU = offset
-#     rgb_frames = video[idx:idx+offset,:,:,:]
-#     print(rgb_frames.shape)
-#     final_img = segment_images_batch(rgb_frames)
-# b=time.time()
-# _secs = (b-a)%60
-
-# print("FPS:",video_len/_secs)
+def process_video(INPUT_FILE,OUTPUT_FILE):
+    video = VideoFileClip(INPUT_FILE)
+    processed_video = video.fl_image(segment_images)
+    processed_video.write_videofile(OUTPUT_FILE, audio=False)
+
+process_video('./test_video.mp4','./output_video.mp4')
 
 exit()