With OpenCV(Python), let's evaluate video similarity!
Given folder, it reads all its own folder. after find images, it starts to convert 32x32 img, RGB Channel.
But this script is unused!
just modify PATH in file, and enjoy how to works!
you need to install [pickle], [PIL], [numpy]
Given folder, it reads all its own folder like [GenBinFile.py].
it makes custom CIFAR-10 data 50000 train data & 10000 valid data.
if you have 60000 data more, don't worry.
when it makes long LIst, it shuffles List, and cuts exactly at 60000.
python GenBinFile2.py
modify PATH correctly.
it is made for only 10 classes. but it supports more. structure is shown below
b'data': R(1,1024), G(1,1024), B(1,1024) with uint 8 b'labels': 0~9 index b'names': index _ imageName
you may have some questions about [GenBinFile2.py] work correctly.
Modify PATH also and enjoy.
python LoadBinFile.py
it shows " b'labels' b'data' b'names'".
Modify PATH and file name that you want to extract.
line 30, you should judge frame variable.
This is sample OpenCV SSIM caculator.
modify img1,2's path. (you must know that exactly same width, height requires)
and it shows how different they are. (it depends on line 28)
it converts video to image. but it doesnt depend on frame(when you define method SSIM, it will use SSIM with tolerance and extract image).
python videoToImage --file aaa.mp4 --method (SSIM, ORB, SIFT, FLANN_SIFT, FLANN_ORB) --debug (Y or N) --tolerance (0.0~1.0) --factor (0.0~1.0)
but ORB,SIFT,FLANN_SIFT,FLANN_ORB is implemented. but i didnt make extract code!
First, It will find starting point(frame) of Original video based on target video(first frame).
Second, it will calculate max SSIM score & frame index.
Third, With starting point(frame), Caculate SSIM again with tolerance. it will devide similar or not.
Fourth, Count number of result. and Decide how similar they are.
python FindMostSimilarityScene.py --o origin.mp4 --t target.mp4 --debug (Y or N) --resize (Y or N)
- resize option makes images 32x32.
it will call a shell (need to be activated anaconda).
it uses pretrained inception_cifar.py.
python Similarity.py -o fortest.mp4