Object Detection

Object Detection model using tensorflow.

1.Install Dependencies

pip install -r requirement.txt

2.Data Collection using opencv

python collect_data.py

Captured images are stored in data/images

3.Annotate Images with LabelMe

labelme

• It pop up a gui .

• Open dir -> data/images

• Change output dir -> data/labels

• Select "Create Rectangle" in edit tab and select the object and give a label name.

4.Split datas to train , test and split.

Datas are distributed among different folder named test , train , val . Each of these folder contains images and labels.

python data_split.py

5.Data augmentation

Datas are augmented using albumentations.The stored data along with labels will store in aug_data folder.

python data_augmentation.py

6.Model Training

Pretrained model vgg16 is used for training. Output of vgg16 is given to a classification model which check wheather a face present and regression model to predict the position of the face. Seperate loss functions are used for classification and regression.

7.Save Model

Model saved as facetrain.h5 .

8.Real time object detection

Opencv-python is used for real time object detection . It shows a rectangle bounding object if it present in the frame.

python real_time.py