How to do real-time testing with a camera?

[ahmetgunduz]

No description provided.

[sanaz97]

Hi Ahmet,
I ran online_test.py and it doesn't have any error and creates 2 json files:

Opt_clf.json
Opt_det.json

can you explain what are these jsons?

and my question is, Is it possible to run your model to test it in real time using webcam?
if it is can you explain the procedure?

I'm testing it with the model that trained by egogesture dataset and I want to test it on my laptap 's webcam

[ahmetgunduz]

Hi @sanaz97 ,

Those jsons are simply for saving the parameters where you specified for classifier and detector architectures. It does not used in the code. You do not need to worry about them unless you want to rerun the test with exactly the same settings you may use them as reference for yourself.

It is possible to run these models with a wabcam but that will require different data loading procedure. Please check opencv specifically cv2.VideoCapture() method. And with a quick search I found this medium blog where it explain how to run a keras model on webcam video. As long as you capture the data from webcam and cast the frames into torch.FloatTensor you can run your model like online_test.py with different data loading.

[sathiez]

Hi,
How to feed input to classifier in online_test.py using tensor.float .
I tried ,

frame= np.reshape(frame,(1,1,1,512,512)) frame=cv2.normalize(frame,None,alpha=0,beta=1,norm_type=cv2.NORM_MINMAX,dtype=cv2.CV_32F)
input_clf = torch.from_numpy(frame).float()
outputs_det = classifier(inputs_clf)

I get the following error,

RuntimeError: invalid argument 2: input image (T: 1 H: 32 W: 16) smaller than kernel size (kT: 2 kH: 3 kW: 3) at /pytorch/aten/src/THCUNN/generic/VolumetricAveragePooling.cu:57

[ahmetgunduz]

@sathiez this is not related with this issue, Please open a new one for it!!.

[ghost]

Is the homepage simulator code available?

[ghost]

@ahmetgunduz Is the home simulator code available?

[ahmetgunduz]

Unfortunately it is not available. What I did back then, using matplotlib simulation plot on class probabilities on a sample video from Egogesture.

[ChunJyeBehBeh]

Hello, may I know how to run the recognition_model with RGB camera? Which code file to run and where should I put the pretrained model?