Marco leonardi is
Main feedback, summed up:
A close to optimal bubble detection can be obtained by retraining the Mask RCNN network. Here is the link to the official github repository: https://github.com/matterport/Mask_RCNN
Traditional pre-processing
Our data set is affected by noise (the cells) and non-contant enivronment parameters such as illumination, shadows, various disturbtions. When this is the case, traditional preprocessing shows very bad performance.
Hence, our alternatives are the following:
- Ensure constant parameters: contants lighting, uniform background, decrease probabilities of shadows, dust etc.
- Model the noise and the change in parameters in order to compensate for these in a more structured and controlled way (rather than relying on built in adaptive thresholds that may or may not work).
- Training of more advanced neural networks
Darknet YOLO (You Only Look Once)
Darknet YOLO is a very common neural network, but the reason why it is so broadly used is because people adapt from others and this one is known as the one "everybody uses " (according to Leonardi).
In our small period of trying this out, we experienced that training this network with the raw images resulted in relatively bad performance. Lots of false positives and few actual detections. However our results improved drastically when the dataset was binarized with edge detection and dilation (hence a fairly heavy and slow preprocessing procedure in advance). According to Leonardi this makes sense as these networks are way easier to train on binary images. However, he was not convinced that the results were optimal.
Mask RCNN vs. Darknet YOLO
There are two main differences between these two neural networks that in Leonardi's opinion would lead to way better performance using RCNN.
- Mask RCNN labels each pixel in the image. YOLO on the other side only tries to find a bounding box around the desired contours. This has proven to give better performance in object detection.
- Mask RCNN is a significantly larger neural network. It is therefore slower in execution (and each training takes longer), but on the other hand there are more filters and the object detection becomes more accurate. The networks is capable of detecting more complex objects than Darknet YOLO.
Challenges with Mask RCNN
A larger and more complex neural network is also harder to understand and adapt to your desired usage. Hence understanding how the network functions and how to retrain the last layers may be challenging. Also, there are fewer people out there with experience and probably less help available in forums etc.
Computational cost vs. speed is always a challenge.
The value of bubble detection vs. the time necessary to implement this into the system must definitely be taken into account before deciding on this approach.