Hugging Face's logo

Spaces:
EdgeFirst
/
README
Running

App Files Community
sebastientaylor commited on
Commit
07c27c7
·
verified ·
1 Parent(s): d9c6f60

Move diagrams and full content into README.md for org page rendering

Browse files
Files changed (1) hide show
  1. README.md +91 -16
README.md CHANGED
@@ -10,28 +10,103 @@ license: apache-2.0
10
 
11
  # EdgeFirst AI — Spatial Perception at the Edge
12
 
13
- Open-source libraries and microservices for AI-driven spatial perception on edge devices.
14
 
15
- **EdgeFirst Perception** supports cameras, LiDAR, radar, and time-of-flight sensors — enabling real-time object detection, segmentation, sensor fusion, and 3D spatial understanding, all optimized for resource-constrained embedded hardware.
 
 
 
16
 
17
- ## Architecture
 
 
 
 
 
 
 
 
 
 
 
 
 
 
18
 
19
- | Layer | Description | Status |
20
- |-------|-------------|--------|
21
- | **Foundation** | Hardware abstraction, video I/O, accelerated inference delegates | Stable |
22
- | **Zenoh** | Modular perception pipeline over Zenoh pub/sub | Stable |
23
- | **GStreamer** | Spatial perception elements for GStreamer / NNStreamer | Stable |
24
- | **ROS 2** | Native ROS 2 nodes extending Zenoh microservices | Roadmap |
25
 
26
  ## Supported Hardware
27
 
28
- NXP i.MX 8M Plus | NXP i.MX 93 | NXP i.MX 95 | NXP Ara240 | RPi5 + Hailo-8/8L | NVIDIA Jetson
 
 
 
 
 
 
 
 
 
 
 
 
29
 
30
- ## Links
 
 
 
 
 
31
 
32
- - [EdgeFirst Studio](https://edgefirst.studio) — MLOps platform (free tier available)
33
- - [GitHub](https://github.com/EdgeFirstAI) — Open-source repositories
34
- - [Documentation](https://doc.edgefirst.ai) — Full documentation
35
- - [Au-Zone Technologies](https://www.au-zone.com) — Company
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
36
 
37
- Apache 2.0 | Au-Zone Technologies Inc.
 
10
 
11
  # EdgeFirst AI — Spatial Perception at the Edge
12
 
13
+ **EdgeFirst Perception** is an open-source suite of libraries and microservices for AI-driven spatial perception on edge devices. It supports cameras, LiDAR, radar, and time-of-flight sensors — enabling real-time object detection, segmentation, sensor fusion, and 3D spatial understanding, optimized for resource-constrained embedded hardware.
14
 
15
+ [![EdgeFirst Studio](https://img.shields.io/badge/EdgeFirst_Studio-3E3371?style=for-the-badge&logoColor=white)](https://edgefirst.studio)
16
+ [![GitHub](https://img.shields.io/badge/GitHub-212529?style=for-the-badge&logo=github&logoColor=white)](https://github.com/EdgeFirstAI)
17
+ [![Documentation](https://img.shields.io/badge/Documentation-1FA0A8?style=for-the-badge&logo=readthedocs&logoColor=white)](https://doc.edgefirst.ai)
18
+ [![Au-Zone Technologies](https://img.shields.io/badge/Au--Zone_Technologies-6C757D?style=for-the-badge)](https://www.au-zone.com)
19
 
20
+ ---
21
+
22
+ ## Workflow
23
+
24
+ ![EdgeFirst Model Zoo Ecosystem](01-ecosystem.png)
25
+
26
+ Every model in the EdgeFirst Model Zoo passes through a validated pipeline. [**EdgeFirst Studio**](https://edgefirst.studio) manages datasets, training, multi-format export (ONNX, TFLite INT8, eIQ Neutron, Kinara DVM, HailoRT HEF, TensorRT), and reference validation. Models are then deployed to our board farm for **full-dataset on-target validation** on real hardware — measuring both accuracy (mAP) and detailed timing breakdown per device. Results are published here on HuggingFace with per-platform performance tables.
27
+
28
+ ## Model Lifecycle
29
+
30
+ ![Model Lifecycle: Training to Publication](02-model-lifecycle.png)
31
+
32
+ ## On-Target Validation
33
+
34
+ ![On-Target Validation Pipeline](03-on-target-validation.png)
35
 
36
+ Unlike desktop-only benchmarks, EdgeFirst validates every model on **real target hardware** with the full dataset. Each device produces both accuracy metrics (mAP) and a detailed timing breakdown — load, preprocessing, NPU inference, and decode — so you know exactly how a model performs on your specific platform.
37
+
38
+ ---
 
 
 
39
 
40
  ## Supported Hardware
41
 
42
+ ![NXP i.MX 8M Plus](https://img.shields.io/badge/NXP-i.MX_8M_Plus-3E3371?style=flat-square&logoColor=white)
43
+ ![NXP i.MX 95](https://img.shields.io/badge/NXP-i.MX_95-3E3371?style=flat-square&logoColor=white)
44
+ ![NXP Ara240](https://img.shields.io/badge/NXP-Ara240-3E3371?style=flat-square&logoColor=white)
45
+ ![RPi5 + Hailo-8/8L](https://img.shields.io/badge/RPi5-Hailo--8%2F8L-1FA0A8?style=flat-square&logoColor=white)
46
+ ![NVIDIA Jetson](https://img.shields.io/badge/NVIDIA-Jetson-76B900?style=flat-square&logoColor=white)
47
+
48
+ ---
49
+
50
+ ## Model Zoo
51
+
52
+ Pre-trained YOLO models for edge deployment. Each model repo contains all sizes (nano through x-large), ONNX FP32 and TFLite INT8 formats, with platform-specific compiled variants as they become available.
53
+
54
+ ### Detection
55
 
56
+ | Model | Sizes | Nano mAP@0.5 | Link |
57
+ |-------|-------|-------------|------|
58
+ | **YOLO26** | n/s/m/l/x | 54.9% | [EdgeFirst/yolo26-det](https://huggingface.co/EdgeFirst/yolo26-det) |
59
+ | **YOLO11** | n/s/m/l/x | 53.4% | [EdgeFirst/yolo11-det](https://huggingface.co/EdgeFirst/yolo11-det) |
60
+ | **YOLOv8** | n/s/m/l/x | 50.2% | [EdgeFirst/yolov8-det](https://huggingface.co/EdgeFirst/yolov8-det) |
61
+ | **YOLOv5** | n/s/m/l/x | 49.6% | [EdgeFirst/yolov5-det](https://huggingface.co/EdgeFirst/yolov5-det) |
62
 
63
+ ### Instance Segmentation
64
+
65
+ | Model | Sizes | Nano Mask mAP | Link |
66
+ |-------|-------|--------------|------|
67
+ | **YOLO26** | n/s/m/l/x | 37.0% | [EdgeFirst/yolo26-seg](https://huggingface.co/EdgeFirst/yolo26-seg) |
68
+ | **YOLO11** | n/s/m/l/x | 35.5% | [EdgeFirst/yolo11-seg](https://huggingface.co/EdgeFirst/yolo11-seg) |
69
+ | **YOLOv8** | n/s/m/l/x | 34.1% | [EdgeFirst/yolov8-seg](https://huggingface.co/EdgeFirst/yolov8-seg) |
70
+
71
+ ---
72
+
73
+ ## Naming Convention
74
+
75
+ | Component | Pattern | Example |
76
+ |-----------|---------|---------|
77
+ | HF Repo | `EdgeFirst/{version}-{task}` | `EdgeFirst/yolov8-det` |
78
+ | ONNX Model | `{version}{size}-{task}-coco.onnx` | `yolov8n-det-coco.onnx` |
79
+ | TFLite Model | `{version}{size}-{task}-coco.tflite` | `yolov8n-det-coco.tflite` |
80
+ | i.MX 95 Model | `{version}{size}-{task}-coco.imx95.tflite` | `yolov8n-det-coco.imx95.tflite` |
81
+ | Studio Project | `{Dataset} {Task}` | `COCO Detection` |
82
+ | Studio Experiment | `{Version} {Task}` | `YOLOv8 Detection` |
83
+
84
+ ## Validation Pipeline
85
+
86
+ | Stage | What | Where |
87
+ |-------|------|-------|
88
+ | **Reference** | ONNX FP32 and TFLite INT8 mAP on full COCO val2017 (5000 images) | EdgeFirst Studio (cloud) |
89
+ | **On-Target** | Full dataset mAP + timing breakdown per device | Board farm (real hardware) |
90
+
91
+ ## Perception Architecture
92
+
93
+ | Layer | Description |
94
+ |-------|-------------|
95
+ | **Foundation** | Hardware abstraction, video I/O, accelerated inference delegates |
96
+ | **Zenoh** | Modular perception pipeline over Zenoh pub/sub |
97
+ | **GStreamer** | Spatial perception elements for GStreamer / NNStreamer |
98
+ | **ROS 2** | Native ROS 2 nodes extending Zenoh microservices *(Roadmap)* |
99
+
100
+ ## EdgeFirst Studio
101
+
102
+ [**EdgeFirst Studio**](https://edgefirst.studio) is the MLOps platform that drives the entire model zoo pipeline. **Free tier available.**
103
+
104
+ - Dataset management & AI-assisted annotation
105
+ - Model training with automatic multi-format export and INT8 quantization
106
+ - Reference and on-target validation with full metrics collection
107
+ - CameraAdaptor integration for native sensor format training
108
+ - Deploy trained models to edge devices via the [EdgeFirst Client](https://github.com/EdgeFirstAI/client) CLI
109
+
110
+ ---
111
 
112
+ Apache 2.0 · [Au-Zone Technologies Inc.](https://www.au-zone.com)