Application of object detection methods state-of-the-art, including YOLO series, mobilenet-SSD, Mask-RCNN up to now.
(The code comments are partly descibed in chinese)
When it comes to deep learning-based object detection there are three primary object detection methods that you’ll likely encounter:
Faster R-CNNs
You Only Look Once (YOLO)
Single Shot Detectors (SSDs)
-
Faster R-CNNs are likely the most “heard of” method for object detection using deep learning; however, the technique can be difficult to understand (especially for beginners in deep learning), hard to implement, and challenging to train. Furthermore, even with the “faster” implementation R-CNNs (where the R stands for Region Proposal), the algorithm can be quite slow, on the order of 7 FPS even when deploying top-level GPU.
Mask-RCNN is a new member of RCNN series, which can not only detect objects but also segment its shape, and it was trained on specific MSCOCO dataset. -
If we are looking for pure speed then we tend to use YOLO as this algorithm is much faster, capable of processing 40-90 FPS on a Titan X GPU. The super fast variant of YOLO can even get up to 155 FPS. The problem with YOLO is that it leaves much accuracy to be desired.
-
SSDs, originally developed by Google, is a balance between the two. The algorithm is more straightforward than Faster R-CNNs. Here we used is SSD based on MobileNet, which simplifies the computation and run much faster to satisfy real-time need but lower the accuracy pretty much meanwhile.
The newest updated version —— YOLOv3, has achieved very comparable accuracy than SSD while running much faster.
In each seperated directory, you can see more details in its README.md.