search

mailrocketsystems / JetsonYolov5

GNU General Public License v3.0
21 stars 22 forks source link

pycuda._driver.LogicError For camera stream #12

Open Jeevi10 opened 4 months ago

Jeevi10 commented 4 months ago

Traceback (most recent call last): File "app.py", line 32, in detections, t = model.Inference(frame) File "/jetson-inference/data/networks/JetsonYolov5/yoloDet.py", line 93, in Inference stream = cuda.Stream() pycuda._driver.LogicError: cuStreamCreate failed: context is destroyed

model runs perfectly with local file and when I feed my live camera I am getting this above error. Can someone guide me about this issue. I appreciate your help

RN0H commented 2 months ago

I am facing the same issue

RN0H commented 1 month ago

Traceback (most recent call last): File "app.py", line 32, in detections, t = model.Inference(frame) File "/jetson-inference/data/networks/JetsonYolov5/yoloDet.py", line 93, in Inference stream = cuda.Stream() pycuda._driver.LogicError: cuStreamCreate failed: context is destroyed

model runs perfectly with local file and when I feed my live camera I am getting this above error. Can someone guide me about this issue. I appreciate your help

I was able to solve this issue, by using pyrealsense2. I created a seperate pipeline for streaming that did not involve cuda context handling. Previously I was subscribing to a rostopic for camera images, and I tried to handle the context, but it was giving me cuda driver errors. I do not know the exact solution to the ros subscriber method. Nevertheless, I am sticking to using pyrealsense2 rather than ros for this one.