HEAD DETECTION: The model was trained (transfer learning + fine-tuning) with TensorFlow Object-Detection API using MobileNetV2 pretrained weights. For classification MobilenetV2 and for detection SSD300 architectures were used. The model is trained on SCUT-HEAD ([ Ссылка ]) datasets. For testing, a short clip of a video ([ Ссылка ]) was used. The complete project can be reached out @ [ Ссылка ].

свернуть