Hello everyone,
I'm working on a computer vision project in which I need to detect in real time a moving object and check whether is still integer or not.
Images will be supplied from a 4K videocamera. The object should be in foreground; I have experience in classification with initial CNNs such as LeNet and pre trained VGG. What architecture/technique is best suited for this kind of problem? I'm worried about fps detection time and amount of data (frames) to train deeper architectures.
Thanks,
Hello,
From the description of your problem, it seems a real-time adaptive learning process. You can think about Deep Reinforcement Learning. May be any deep network with Q learning should be a good approach.
can be done in matlab using kalman filter
https://in.mathworks.com/support/search.html/videos/introduction-to-kalman-filters-for-object-tracking-79674.html
- Badges
- Science topic
- Similar topics
- Design
- Architecture