Spaces:
Running
Running
Refresh: 789 validation sessions, 20 hardware platforms; humanizer pass on published prose
Browse files
README.md
CHANGED
|
@@ -18,7 +18,7 @@ Each model family lives in its own HuggingFace repo containing all size variants
|
|
| 18 |
|
| 19 |
---
|
| 20 |
|
| 21 |
-
## Model
|
| 22 |
|
| 23 |
### Detection
|
| 24 |
|
|
@@ -39,7 +39,7 @@ Each model family lives in its own HuggingFace repo containing all size variants
|
|
| 39 |
|
| 40 |
---
|
| 41 |
|
| 42 |
-
## Repo
|
| 43 |
|
| 44 |
Each model repo follows a consistent layout with platform folders:
|
| 45 |
|
|
@@ -64,7 +64,7 @@ EdgeFirst/yolov8-det/
|
|
| 64 |
└── ...
|
| 65 |
```
|
| 66 |
|
| 67 |
-
## Naming
|
| 68 |
|
| 69 |
**Pattern**: `{version}{size}-{task}-{precision}[-{variant}][.{platform}].{ext}`
|
| 70 |
|
|
@@ -91,7 +91,7 @@ EdgeFirst/yolov8-det/
|
|
| 91 |
| Hailo-8L HEF | `yolov8n-det-int8.hailo8l.hef` |
|
| 92 |
| Jetson TensorRT FP16 | `yolov8n-det-fp16.orin-nano.engine` |
|
| 93 |
|
| 94 |
-
## Supported
|
| 95 |
|
| 96 |
       
|
| 97 |
|
|
@@ -104,7 +104,7 @@ EdgeFirst/yolov8-det/
|
|
| 104 |
- **RPi5 + Hailo-8L** — 13 TOPS, HailoRT HEF
|
| 105 |
- **NVIDIA Jetson Orin** — 67–157 TOPS, TensorRT
|
| 106 |
|
| 107 |
-
## Validation
|
| 108 |
|
| 109 |
Every artifact in the Model Zoo is measured on the same dataset on the same hardware users deploy on. Accuracy numbers and per-stage timing are produced by the same pipeline that runs the deployed model — there is no "benchmark configuration" separate from production.
|
| 110 |
|
|
|
|
| 18 |
|
| 19 |
---
|
| 20 |
|
| 21 |
+
## Model repositories
|
| 22 |
|
| 23 |
### Detection
|
| 24 |
|
|
|
|
| 39 |
|
| 40 |
---
|
| 41 |
|
| 42 |
+
## Repo structure
|
| 43 |
|
| 44 |
Each model repo follows a consistent layout with platform folders:
|
| 45 |
|
|
|
|
| 64 |
└── ...
|
| 65 |
```
|
| 66 |
|
| 67 |
+
## Naming convention
|
| 68 |
|
| 69 |
**Pattern**: `{version}{size}-{task}-{precision}[-{variant}][.{platform}].{ext}`
|
| 70 |
|
|
|
|
| 91 |
| Hailo-8L HEF | `yolov8n-det-int8.hailo8l.hef` |
|
| 92 |
| Jetson TensorRT FP16 | `yolov8n-det-fp16.orin-nano.engine` |
|
| 93 |
|
| 94 |
+
## Supported hardware
|
| 95 |
|
| 96 |
       
|
| 97 |
|
|
|
|
| 104 |
- **RPi5 + Hailo-8L** — 13 TOPS, HailoRT HEF
|
| 105 |
- **NVIDIA Jetson Orin** — 67–157 TOPS, TensorRT
|
| 106 |
|
| 107 |
+
## Validation pipeline
|
| 108 |
|
| 109 |
Every artifact in the Model Zoo is measured on the same dataset on the same hardware users deploy on. Accuracy numbers and per-stage timing are produced by the same pipeline that runs the deployed model — there is no "benchmark configuration" separate from production.
|
| 110 |
|