Human activity recognition strives to identify activities from a series of observations on the actions of subjects and the environmental conditions. Here, we have recognized certain activities such as yoga, handwashing, cycling by training our model accordingly.
- We trained our model with various activity images
- With help of Transfer Learning, we can identify human’s visual-skeleton
- For the classification of different actions, the CNN model was trained
- Once a human is identified, it predicts its pose using Computer Vision, based on visual-skeleton
- Later on, it predicts its behavior by analyzing human’s behavior according to the trained model
Key values / Achievements
- The output shows the predicted scene with score.
- The model was able to identify the human visual skeleton and based upon that it accurately detected the scene with prediction score.
- Deep learning and transfer learning helped to achieve good accuracy.