GitHub - paolo-moriello/Automatic-Video-Shot-Type-Classification: ResNet model in TensorFlow

Branches Tags

Name		Name	Last commit message	Last commit date
Latest commit History 173 Commits
Graphs		Graphs
.gitignore		.gitignore
bright_image.py		bright_image.py
copy_audio_from_video.sh		copy_audio_from_video.sh
crop_image.py		crop_image.py
crop_image_grid.py		crop_image_grid.py
extract_frames.sh		extract_frames.sh
flip_image.py		flip_image.py
get_default_frames.py		get_default_frames.py
get_default_ratio.py		get_default_ratio.py
get_video_info.sh		get_video_info.sh
graph_github.txt		graph_github.txt
graph_hub.txt		graph_hub.txt
lenet.py		lenet.py
move_images.sh		move_images.sh
outputGP		outputGP
outputPM		outputPM
outputPR		outputPR
parse_video_info.py		parse_video_info.py
pred.csv		pred.csv
readme.txt		readme.txt
retrain_last_layer.py		retrain_last_layer.py
retrain_last_layer_lenet.py		retrain_last_layer_lenet.py
run_all.sh		run_all.sh
run_crop.py		run_crop.py
run_extract_frame.sh		run_extract_frame.sh
run_extract_frames.sh		run_extract_frames.sh
scale_image.py		scale_image.py
scnn.py		scnn.py
scnn_mod.py		scnn_mod.py
test_model.py		test_model.py
utils.py		utils.py
write_predictions_video.py		write_predictions_video.py

Repository files navigation

After having installed all the libraries needed, running the resnet is very simple: just run the forward.py file inside tensorflow, using python:
- (tensorflow) python forward.py

NB: the forward.py has a parameter inside which specify the length of the net for usage of pretrained models.

Our scripts:

- crop_image.py : given the img name and an amount of pixels, it produces 4 images, each one with a cropped size of the given amount
- run_crop.py : script to run the crop_img_grid script which produce pixels*pixels images (python run_crop.py /datas/teaching/projects/spring2018/ps34/Data/Images_Plans/)
- extract_frames.sh : bash script to extract frames from video (parameters: video path + list frame number)