Hello , i need help regarding using tensorflow v1 model on zed camera . I have trained my model using custom data , but it is fluctuating a lot. Displaying random boxes on screen
Hi @SUJAY_MHASKE,
I apologize for the late reply. You may have already found a solution, but here are a few suggestions. Based on the TensorFlow documentation, the V1 Object Detection API is no longer actively maintained. However you can refer the actively maintained codebase here for your usecase. Please make sure that the ZED SDK version is compatible with Tensorflow 2.X.
The fluctuations while inferencing might be due to input frame rate is more compared to the model’s inference speed and make sure input frame size should match with image size.
Thank You