Open mengniwang95 opened 3 years ago
I found there are about 11 models without accuracy or preprocess details in cv category. For those models without preprocess details, if I use some typical preprocess but get different accuracy, is it acceptable? For those models without accuracy, I can't confirm whether my results are correct, is it acceptable? @wenbingl
When the preprocessing of the inception model is the same as resnet, the inference result is wrong inception v1: resnresnet18-v1:
how to solve?
update model list which misses preprocess/postprocess details: https://github.com/onnx/models/tree/main/vision/classification/rcnn_ilsvrc13 https://github.com/onnx/models/tree/main/vision/classification/inception_and_googlenet/inception_v2 https://github.com/onnx/models/tree/main/vision/object_detection_segmentation/tiny-yolov2 https://github.com/onnx/models/tree/main/vision/object_detection_segmentation/yolov2-coco https://github.com/onnx/models/tree/main/vision/body_analysis/ultraface https://github.com/onnx/models/tree/main/text/machine_comprehension/gpt-2
I found there are some models lack data preprocess details and even accuracy info, such as https://github.com/onnx/models/tree/master/vision/classification/inception_and_googlenet/inception_v2 Where can I get these details to reproduce results?