Dr. Uwe Köhler

Still doesn't work in OpenCV 4.7.0-dev

Thanks for letting me know!

This is still an issue in for opencv-python-rolling 4.7.0.20230211 models: fcn-resnet101-11.onnx ssd-12.onnx produces a different error now (https://github.com/opencv/opencv/issues/23298) FasterRCNN-12.onnx produces a different error now (https://github.com/opencv/opencv/issues/23297)

Possible workaround (at least infer does something for fcn-resnet101-11.onnx). @fengyuentau pointed me to the possibility to fix the input size of an image in an ONNX model using onnxsim (https://github.com/opencv/opencv/issues/23288)...

After a paddlehub up date the install works now: ``` (.venv) D:\Local\devel\Python\PaddleSeg_2.7.0>hub install deeplabv3p_xception65_ humanseg Download https://bj.bcebos.com/paddlehub/paddlehub_dev/deeplabv3p_xception65_hum anseg_1_2_0.zip [##################################################] 100.00% Decompress C:\Users\ukoehler\.paddlehub\tmp\tmptznv48kc\deeplabv3p_xception65_hu manseg_1_2_0.zip [##################################################] 100.00% [2023-01-10 12:09:39,796] [ INFO] -...

The update got me a bit further. The ONNX export does now fail, however: https://github.com/PaddlePaddle/Paddle2ONNX/issues/1008

Problem confirmed for super-gradients 3.1.1 Platform Windows 10 from super_gradients.training import Trainer is enough to cause the problem for me. Workaround sys.stdout = sys.__stdout__ works, but it is still annoying.

Same problem here. @haritsahm did help a lot, However, the boxes are slightly shifted. At least it does prove that the basic information is available. Now, the question is: Is...

Hmmm, I followed a tutorial to trained starting from pre-trained data. The question is: when the input images are scaled in an unsuitable way for the pre-trained dataset (YOLO-NAS), is...

Hi @iAlokMishra, the poor performance is due to image normalization (or not). See @haritsahm solution/work-around.