Excuse me! I have some questions when I use pot(Post-Training Optimization Tool) command. It required to input dataset and the dataset formats are ImageNet, Pascal VOC, COCO, Common Semantic Segmentation, Common Super-resolution, LFW, VGGFaces2, WIDER FACE, Open Images, or Not Annotated. If my data dimension is [1, 16, 8, 10], how can I input my data?
Thanks for reaching out to us.
To use the Post-Training Optimization Tool (POT) you need a floating-point precision model, FP32 or FP16, converted into the OpenVINO Intermediate Representation (IR) format. The POT accepts models in IR format only and you need to convert your model into the IR format using Model Optimizer.
Thus, for the input data, you will need to specify it during the conversion into the IR format using Model Optimizer. Run Model Optimizer with the --input_shape option to specify full shapes for each input data.
Regarding the usage of the input dataset, you will need to use POT API for the implementation. There are specific interfaces such as Engine, DataLoader and Metric that will call the custom optimization pipeline. Check out the Post-Training Optimization Tool API documentation for a detailed explanation.
This thread will no longer be monitored since we have provided a solution. If you need any additional information from Intel, please submit a new question.