How to enable object detection Tensorflow api to use gray image for training (only 1 channel for input tensor)?
I just want to get real time speed when using a model trained by tensorflow api object detection, the input tensor is of the shape [1, width, height, 3], these are 3 channels, but I think if I can just use 1 channel for training my model, it just needs grayscale images as input, so this can reduce the computational complexity, which is very important for my application.
source to share
We do not have pre-trained models that render single-channel images. If you are interested in creating a new model, consider adding a new SSD Feature Extractor .
Speaking of which, you may not see any tangible acceleration in your model. Using only one channel only affects the input layer of the network, not intermediate layers. Better to consider using the more subtle Mobilenet feature extractor if performance is your biggest concern.
source to share