Result
The final result of our project is a system that would allow user to control the light using the gesture. In this section, we will display our images taken by the hand detection camera, and the analysis result of these images by using the gesture detecting and finger counting logic. Using these results and comparing them with the one-to-one match between gesture and light set in advance in the code(for example thumb up to green light), we can change the voltage levels of light pins and eventually change the lights.
The following pictures is the images taken by hand detection with analysis result.
Figure 1: Resulting image
Figure 2: Desired Result
Figure 3: Original image
Figure 4: Image after binary mask
Discussion of Results
The quality of the hand detection was not very ideal. As the figure shows there are many other pictures besides the hand that are detected. That is probably because of the dataset is made by ourselves and not fit enough for training the detection. If we have more time, we could have attempted to make or find a more reliable dataset and took more time to train the model.
Thanks to the robustness of the system, the wrong detection will make few influence of the final result. In the process of finger counting, it will convert a binary mask to the image, and then use the logic to detect the number of finger. For the wrong detection, the image of something will be put into the finger counting process and be detected as invalid in theory.
In practice, the binary mask of finger is not perfect. The light and shadow will influence the result of the binary mask, and influence the quality of finger counting also. We should improve the method for masking to increase make a more stable finger counting.