Rafael Escarabajal

dinov3_ros. ROS 2 integration of Meta’s DINOv3 backbone with lightweight heads for vision tasks.

dinov3_deepstream. DeepStream integration of Meta’s DINOv3 backbone with lightweight heads for vision tasks.

depth_dinov3. Lightweight head for depth estimation using DINOv3 as backbone

depth_anything_v3_deepstream. Real-time monocular depth estimation with Depth Anything V3 and NVIDIA DeepStream 8.0.

object_detection_dinov3. Lightweight head for object detection using DINOv3 as backbone

semantic_segmentation_dinov3. Lightweight head for semantic segmentation using DINOv3 as backbone

optical_flow_dinov3. Lightweight head for optical flow using DINOv3 as backbone

Radar-camera-fusion. Code to perform radar and camera fusion with a NN and the NuScenes dataset. Lidar is used for supervision

tensorrt_lib. Simple library of TensorRT for C++ and Python with use cases

siam_tracking. Code to perform single-object tracking using siamese tracker.

reconstruction_3d. Code to perform 3D reconstruction using a Realsense camera