$ show Intellindust-AI-Lab/DEIMv2
DEIMv2 is a production-grade object detection framework built upon the DETR architecture, specifically implementing the DEIM (DETR with Improved Matching) algorithm extended with DINOv3 features. The codebase provides a modular design with multiple backbone options (HGNetv2, DINOv3, ResNet variants), hybrid encoders, and transformer decoders (DEIM, D-FINE, RT-DETRv2). It includes comprehensive training scripts, YAML-based configuration inheritance, data augmentation pipelines, and deployment tools for ONNX, TensorRT, and OpenVINO. The repository directly trains and evaluates on standard detection benchmarks like COCO and Objects365, and is structured for reproducibility and extensibility. While heavily inspired by prior works (RT-DETR, D-FINE), the implementation demonstrates solid engineering with clean abstractions, distributed training support, and a well-organized module hierarchy.
The LLM identified 3 files worth highlighting for reuse:
Implements the DINOv3 ViT architecture with support for register tokens, RoPE (Rotary Position Embedding), and flexible patch sizes. The module is self-contained (imports only from torch and sibling layers) and can be extracted as a standalone vision transformer backbone for other architectures.
A hybrid encoder that fuses transformer attention with CSPNet-style convolutional blocks, designed for efficient multi-scale feature extraction. The module is isolated, taking only feature maps as input and returning encoded features, making it reusable in any detection pipeline requiring a backbone-agnostic encoder.
A clean, self-contained implementation of Exponential Moving Average (EMA) for model weights, supporting both single-GPU and distributed training. The module can be used to stabilize training of any PyTorch model and is easily extracted for reuse.