It is about detection (bounding boxes) and semantic segmentation. Need to be done in C++ with CMakelists binding, Linux environment. For detection, I used the HoG feature approach and trained the SVM classifier for the purpose of finding objects and the bounding boxes around them. Some tuning is needed because the results I get on the test image are currently meaningless. Scripts for training and creating a new dataset for SVM training purposes have been created. Also, the script related to the formation of a CNN-network for segmentation with the help of transfer-learning was developed in Python. Accuracy on the validation set was high, therefore it is only necessary to import the already trained TensorFlow model in C++ and create a test script.
14 freelancers estão ofertando em média $224 nesse trabalho
HI. I have good experience with c++ object detection and recongnition using opencv or other 3rd party. So hope more disucss with you. If u choose me, you will get best result. Regards.