#163163rd of 168 in ML frameworks and libraries
deim
Training framework for real-time DETR object detectors with faster convergence
- Stars
- 1.7k
- License
- custom license
- Last commit
- Mar 2026
- Language
- Python
1,473 on 21 Apr 20261,650 on 10 Nov 2026
From GitHub's daily star counts.
Checked on GitHub.
Overview
DEIM is the CVPR 2025 reference code for a DETR training method that improves query matching (Dense O2O) so detectors converge faster. It trains and evaluates D-FINE and RT-DETRv2 based models in sizes N to X on COCO or custom COCO-format datasets. Tooling covers ONNX and TensorRT export, inference scripts, latency benchmarks and FiftyOne visualization.
Who it is for: Computer vision engineers training or deploying real-time object detectors
Strengths
- COCO results published: DEIM-D-FINE N 43.0 AP at 4M params, X 56.5 AP at 62M
- Exports to ONNX and TensorRT, with onnxruntime, TensorRT and torch inference scripts
- Custom COCO-format datasets, input size and batch size handled through YAML configs
- Also re-implements D-FINE and RT-DETR, with checkpoints for each size
Weaknesses
- Object detection only; not an LLM or general-purpose AI tool
- Checkpoints are hosted on Google Drive, not a model hub
- Training examples assume 4 GPUs via torchrun; single-GPU guidance is thin
- License metadata is NOASSERTION while the README badge says Apache 2.0
What it needs
- GPU required
- Needs PyTorch, conda, Python 3.11.9, onnx, onnxsim, TensorRT (optional), FiftyOne (optional)
- Models: DEIM-D-FINE (N, S, M, L, X), DEIM-RT-DETRv2 (S, M, M*, L, X), COCO-format datasets, ONNX, TensorRT
Quick answers
Is deim open source?
Its code is public on GitHub under its own license terms; read them before you use it at work.
Does deim need a GPU?
Yes, it needs a GPU.
Also in ML frameworks and libraries
See all 168| Rank | Project | Score |
|---|---|---|
| 1 | transformersPython library defining open models for text, vision, audio and multimodal tasks | 94 out of 100 |
| 2 | pytorchPython tensor library with GPU support and tape-based autograd | 93 out of 100 |
| 3 | ultralyticsPython library and CLI for training and running YOLO vision models | 93 out of 100 |
| #4 | pytorch-image-modelsPyTorch library of image models, pretrained weights and training scripts | 93 out of 100 |
| #5 | supervisionPython toolkit for loading, annotating and counting computer vision detections | 92 out of 100 |