Error 002667 encountered while performing deep learning detection of targets on arcgispro version 3.0.2 with a 3060 graphics card

[leijiang1122]

Deep-learning-frameworks, CUDA11.2, cuDNN11.0 are correctly installed Environment is win11, 3060laptop6G, 16gRAM, I7 12700H ArcGISPro3.0.2 It is worth noting that it runs properly when training the deep-learning model

[leijiang1122]

@scdub I look forward to hearing from you!

[ezw21]

Hi @leijiang1122 , can you try cloning another conda environment with the deep learning essentials downgraded to 2.8
dependencies? That might helps.

[scdub]

@ezw21 In most cases, this will cause issues as the package sets are tied to the product versions in terms of testing and what the tool expects.

@leijiang1122 Can you try rerunning the tool at a location with a short path with only ASCII characters? It is possible one of the tiers of the stack is failing to be Unicode aware and not correctly passing the inputs through the system. If that fails, it at least tells us to look elsewhere.

[leijiang1122]

002667 has been resolved, but there is a new problem error999999

[scdub]

@leijiang1122 Please report this issue to Esri support, unfortunately the tool internals aren't visible from this traceback, but I do see other issues logged related to this.