Abstract: We propose the gradient-weighted Object Detector Activation Maps (ODAM), a visual explanation technique for interpreting the predictions of object detectors. Utilizing the gradients of ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results