I have a dataset of video. I have extracted its frames and pass these frames to ResNet-50. I got 179*7*7*2048 features, I flatten it and get 170*100325 features, and then I normalized all the features.
My ResNet-50 model set is:
model=ResNet50(weights='imagenet', include_top=False, input_shape=(224, 224, 3))
Note: I have to perform anomaly detection on these videos.
But how can I determine that my features extracted are good?
Regards
After extraction of video frame features, you must pass into some classifiers, like SVM, DT, RF, etc. And compute the confusion matrix; if your accuracy and other parameters derived from the confusion matrix are giving good results, it means your extracted features are useful; otherwise, try some different techniques to extract features.
- Badges
- Science topic
- Similar topics
- Phytochemistry
- Extracts