Hi Armaan......Are you Director Priyadarshan's son?
I followed your TF2 OD API tutorial and it was indeed well structured and well explained. Kudos to you!!!
When I used your "TF-video-object-counting.py" and "TF-video-od.py" for inferencing using GPU......
"TF-Video-object-counting.py" counts all the class of objects and shows in the top-left corner and "TF-video-od.py" was clearly differentiating between each class of object with different colors. Both the codes performs these tasks in addition to Object_Detection BB's.....
Attached Powepoint file, shows the details, what I mentioned above.....
I have a use case, where there should be minimum latency while inferencing......The use case is in Industrial Computer Vision and the model has to identify between Real Defects and Pseudo Defects with very minimum latency......The Camera captures the images (which takes about 8 seconds and inference must be done within 1-2 seconds of latency).......I do work on Industrial Edge devices for deployment......
But when I train my models, I would like to check the inferencing using my GPU, as to what FPS I get when running a Video file......Once I am satisfied with my FPS, then I have to convert the model to Intel OpenVINO IR (Intermediate Representation) and deploy the model for real time evaluation.....My Edge device has Intel Movidius Myriad VPU (Vision Processing Unit), which is good for inferencing at the edge.......
I would like to use your TF-video-object-counting.py and TF-video-od.py......It has the best of both worlds......I get the object count in one and in the other one, I could clearly see different colors being used for each class.......And in addition, if I am able to see the "FPS", it would be really helpful for my use case.....
If you can go through the "detect_objects.py" code and make those changes in your code, so that I get one single code with which I can run my model for inferencing to check on FPS, Object count per class and Different colors of Object Detection BB's, I will be grateful to you......
I am a Mechanical Engineer by profession and a novice in coding......Can understand the codes, but when it comes to coding, I can't do by myself........
Hi Armaan......Are you Director Priyadarshan's son?
I followed your TF2 OD API tutorial and it was indeed well structured and well explained. Kudos to you!!!
When I used your "TF-video-object-counting.py" and "TF-video-od.py" for inferencing using GPU......
"TF-Video-object-counting.py" counts all the class of objects and shows in the top-left corner and "TF-video-od.py" was clearly differentiating between each class of object with different colors. Both the codes performs these tasks in addition to Object_Detection BB's.....
Before your tutorial, I used another Python code from this gitbub page..... https://github.com/abdelrahman-gaber/tf2-object-detection-api-tutorial/blob/master/detect_objects.py What the above code "detect_objects.py" gives me in addition to Object Detection BB's are the FPS (Frame Rate Per Second)....
Attached Powepoint file, shows the details, what I mentioned above.....
I have a use case, where there should be minimum latency while inferencing......The use case is in Industrial Computer Vision and the model has to identify between Real Defects and Pseudo Defects with very minimum latency......The Camera captures the images (which takes about 8 seconds and inference must be done within 1-2 seconds of latency).......I do work on Industrial Edge devices for deployment......
But when I train my models, I would like to check the inferencing using my GPU, as to what FPS I get when running a Video file......Once I am satisfied with my FPS, then I have to convert the model to Intel OpenVINO IR (Intermediate Representation) and deploy the model for real time evaluation.....My Edge device has Intel Movidius Myriad VPU (Vision Processing Unit), which is good for inferencing at the edge.......
I would like to use your TF-video-object-counting.py and TF-video-od.py......It has the best of both worlds......I get the object count in one and in the other one, I could clearly see different colors being used for each class.......And in addition, if I am able to see the "FPS", it would be really helpful for my use case.....
If you can go through the "detect_objects.py" code and make those changes in your code, so that I get one single code with which I can run my model for inferencing to check on FPS, Object count per class and Different colors of Object Detection BB's, I will be grateful to you......
I am a Mechanical Engineer by profession and a novice in coding......Can understand the codes, but when it comes to coding, I can't do by myself........
Looking forward to your help and support......
Help_from_Armaan.pptx