stream_infer.py

[idata]

I'm trying to run ../Caffe/SSD_MobileNet with stream_infer.py

Error message:

index 1000 is out of bounds for axis 0 with size 707Output thread terminating

I am having problems I think with SSD_MobileNet >> categories.txt & stat.txt

For stat.tx I got:

0.40787054 0.45752458 0.481093780.00392157 0.00392157 0.00392157

But I think it is wrong

Can anybody advice?

OrHas good link which refers to the Movidius NC Toolkit User Guide and provides the tools and examples to create these files?OrHas anybody got working stat numbers? With correct Categories.txt?

Many Thanks in advance for Helping

[idata]

Hi @VictoryKnocks

The "index 1000 is out of bounds for axis 0 with size 707" error is given because mobilenet SSD is trained on only 20 classes, while others (like squeezenet and googlenet) are trained on about 1000. It would be easiest to modify mobilenetSSD to read from a video instead of an image. SeePlease let me know if you have any further questions!

Regards,

Sahira

[idata]

Hi @Sahira_at_Intel

So I could just change line 199 infer_image = cv2.imread(IMAGE_FULL_PATH)to 199 infer_image = cv2.VideoCapture(0)

And this would capture USB webcam?

[idata]

@Sahira_at_Intel

I run above edit and get this error:

Traceback (most recent call last):

File "run2.py", line 216, in sys.exit(main()) File "run2.py", line 203, in main run_inference(infer_image, graph) File "run2.py", line 38, in run_inference resized_image = preprocess_image(image_to_classify) File "run2.py", line 161, in preprocess_image img = cv2.resize(src, (NETWORK_WIDTH, NETWORK_HEIGHT))TypeError: src is not a numpy array, neither a scalar

[idata]

@Sahira_at_Intel

I can edit video_objects.py to allow me to use usb webcam with ssd_mobilenets instead of video mp4 file

However editing run.py in ssd_mobilenets folder in same way gives error

What am I doing wrong?