My model needs to input 5 consecutive frames of images and detect moving targets in the picture, which uses some 3D convolution. Now it can be successfully converted to the ONNX model, but when converting the RKNN model, an error occurs saying that the input size is incorrect. It seems that it only supports image input. The official yolo example only processes a single image and inputs 4-dimensional data.
Is there any way to make RKNN support 5-dimensional data? (Add one more time dimension on the basis of B, C, H, and W)
My model needs to input 5 consecutive frames of images and detect moving targets in the picture, which uses some 3D convolution. Now it can be successfully converted to the ONNX model, but when converting the RKNN model, an error occurs saying that the input size is incorrect. It seems that it only supports image input. The official yolo example only processes a single image and inputs 4-dimensional data.
Is there any way to make RKNN support 5-dimensional data? (Add one more time dimension on the basis of B, C, H, and W)