I will do instance segmentation and object detection

Core Detection Expertise

• Zero-shot detection using CLIP, DINO, Florence, PaLI-Gemma 2
• LoRA/QLoRA fine-tuning, prompt tuning, visual grounding (Grounding DINO)
• YOLOv5/v8, RetinaNet (single-stage), Faster/Mask R-CNN + FPN (two-stage)
• DETR, DINOv2 with Swin/ViT/SwiftFormer backbones
• Few-shot: Transfer learning, distillation, self-training, active learning

Advanced Tracking & Re-ID

• ByteTrack, DeepSORT (custom embeddings), StrongSORT, Norfair
• Occlusion handling, multi-camera re-ID, Kalman filters, cosine similarity
• Trajectory prediction and association via IoU + Hungarian algorithm

️ Frameworks & Tools

• MMDetection, Detectron2, TF-OD API, Ultralytics, Roboflow and python
• PyTorch, TensorFlow/Keras, OpenCV, Albumentations, Supervision
• Tooling: FiftyOne, W&B, Label Studio, Hydra, MLFlow

Optimization & Deployment

• Triton Server, TensorRT, TorchScript, ONNX, OpenVINO
• Jetson Nano/Xavier, DeepStream, YOLO-TAO, QAT/pruning
• MongoDB, ElasticSearch, FAISS, ChromaDB

Domains

• Industrial: PPE, defect detection
• Surveillance: Threat detection, behaviour modelling
• Transport: ALPR, vehicle tracking

Contact before placing your order