What is the best model for real-time object detection on mobile?

Based on the list, models like YOLO v2 and SSD are highlighted for speed, but compare FLOPs and mAP scores—MobileNet-based detectors might offer better efficiency for edge devices.

How do EfficientNet and ResNet compare for image classification?

EfficientNet models generally provide better parameter efficiency and accuracy scaling; for example, EfficientNet-B0 has a Top-1 Error of 24.77% with 5.3M parameters, while ResNet-50 has 22.28% with 25.5M parameters.

How to choose a segmentation model for high mIOU on Cityscapes?

Check the segmentation table for models with high Cityscapes mIOU scores, like PSPNet (80.2) or DANet (81.5), but consider computational costs and whether the model fits your deployment constraints.

Is this repository updated with the latest Vision Transformer models?

It includes some Vision Transformers like T2T-ViT from 2021, but may lack newer variants (e.g., Swin Transformers). Always cross-reference with recent papers and libraries for up-to-date information.

What does mAP mean in object detection models?

mAP (mean Average Precision) measures detection accuracy across different IoU thresholds; higher mAP indicates better performance, as shown in the detection table for models like Faster R-CNN or RetinaNet.

Awesome Computer Vision Models

CC-BY-4.0

A curated list of popular deep learning models for image classification, segmentation, and detection with key performance metrics.

GitHub

What is Awesome Computer Vision Models?

Awesome Computer Vision Models is a curated GitHub repository that aggregates and organizes popular deep learning models for image classification, segmentation, and object detection. It provides a structured reference with key performance metrics like parameter counts, FLOPs, and accuracy scores, helping researchers and engineers quickly compare and select models for their projects.

Target Audience

Computer vision researchers, machine learning engineers, and students who need a consolidated reference for model architectures and their performance on standard benchmarks like ImageNet, COCO, and Pascal VOC.

Value Proposition

It saves significant research time by compiling hundreds of models with consistent metrics in one place, offering a clear historical and technical overview that is often scattered across papers and blogs.

Overview

A list of popular deep learning models related to classification, segmentation and detection problems

Use Cases

Best For

Comparing model efficiency (parameters vs. accuracy) for deployment

Related Projects

Community-curated · Updated weekly · 100% open source

Found a gem we're missing?

Open-Awesome is built by the community, for the community. Submit a project, suggest an awesome list, or help improve the catalog on GitHub.

Submit a project Star on GitHub

542 stars94 forks0 contributors

Researching the evolution of computer vision architectures

Selecting a baseline model for image classification tasks

Finding segmentation models with high mIOU scores

Identifying real-time object detectors for embedded systems

Studying performance metrics across different datasets and years

Not Ideal For

Teams needing ready-to-use code or pre-trained models for immediate deployment
Researchers requiring the most up-to-date model performances on cutting-edge benchmarks beyond 2021
Practitioners looking for implementation tutorials, training scripts, or deployment guides
Projects focused on niche computer vision tasks not covered by standard classification, segmentation, or detection benchmarks

Pros & Cons

Pros

Structured Model Catalog

Organizes hundreds of models into clear tables for classification, segmentation, and detection, as shown in the README's separate sections with comparative data.

Comparative Performance Metrics

Includes key metrics like parameters, FLOPs, and accuracy scores (e.g., Top-1 Error, mAP) from papers, allowing direct efficiency comparisons across models.

Historical Progression Insight

Tracks evolution from foundational works like AlexNet (2014) to modern architectures like Vision Transformers (2021), providing context for model development trends.

Direct Paper References

Each model entry links to its seminal research paper, facilitating deeper study and verification of metrics, as evidenced by arXiv URLs in the tables.

Cons

Incomplete and Inconsistent Data

Many entries have missing metrics (marked with '?'), such as FLOPs for VGG-16 or specific error rates, which reduces comparability and reliability.

Static and Potentially Outdated

As a static GitHub repository, it may not include models or benchmarks beyond 2021, with no indication of regular updates, requiring users to supplement with current research.

Lacks Practical Implementation

No code examples, training scripts, or deployment guides are provided; users must find implementations elsewhere, limiting immediate utility for hands-on projects.

Frequently Asked Questions

Home

Data Science

vinta

An opinionated list of Python frameworks, libraries, tools, and resources

Stars300,453

Forks28,003

Last commit1 day ago

awesome-machine-learning

A curated list of awesome Machine Learning frameworks, libraries and software.

Stars72,631

Forks15,475

Last commit19 days ago

awesome-awesomeness

A curated list of awesome awesomeness

Stars33,447

Forks3,598

Last commit2 years ago

data-science-ipython-notebooks

Data science Python notebooks: Deep learning (TensorFlow, Theano, Caffe, Keras), scikit-learn, Kaggle, big data (Spark, Hadoop MapReduce, HDFS), matplotlib, pandas, NumPy, SciPy, Python essentials, AWS, and various command lines.

Stars29,138

Forks8,031

Last commit2 years ago

#machine-learning-algorithms

#deep-learning

#neural-networks

#machine-learning-models

#awesome-list

#semantic-segmentation

#model-zoo

#image-classification

#research