this is my cv:
Algorithm Engineer Baidu (China) Co., Ltd.
§ Perform lesion recognition segmentation on medical aspects such as CT and ultrasound images.
§ Use Mask-RCNN, Unet and other algorithms to segment and detect lesions such as thyroid and skin diseases. For high-resolution input images, extract patches, do data enhancement, etc., and apply the stem block network structure to improve the anchor compression ratio. Feature map 1*1 corresponds to the original image pixel 4*4 or even smaller, which is beneficial for very small object detection, such as microcalcification white point; using cascade RCNN, the detector gradually increases the IoU threshold for proposal resampling; finally, do post-processing clustering, convex hull, etc. , to analyze the detected white point. Solve anomalous problems, such as the removal of non-target objects on the outer edge, using traditional image methods, rule removal, etc.
§ Use random generation or cycle gan to learn the noise gan model and use this model to generate white point as the target mask label on the target area.
§ When the data annotation is lacking, the data and model of the similar domain image are migrated, and the graph-cut algorithm is used to do the interactive segmentation annotation. And integrate processing data sets with different distributions.
§ Individual identification of racing pigeons using bio-iris and vein identification techniques; detect the eye area of pigeon images using Faster-Rcnn