sebastientaylor commited on
Commit
33d29ed
·
verified ·
1 Parent(s): e61978d

Update model card for yolov5-det

Browse files
Files changed (1) hide show
  1. README.md +15 -83
README.md CHANGED
@@ -80,13 +80,11 @@ Full pipeline timing: pre-processing + inference + post-processing.
80
 
81
  ## Downloads
82
 
83
- ### Universal Formats
 
84
 
85
- <details>
86
- <summary><strong>ONNX FP32</strong> — Any platform with ONNX Runtime</summary>
87
-
88
- | Size | File | Download |
89
- |------|------|----------|
90
  | Nano | `yolov5n-det-coco.onnx` | [Download](https://huggingface.co/EdgeFirst/yolov5-det/resolve/main/onnx/yolov5n-det-coco.onnx) |
91
  | Small | `yolov5s-det-coco.onnx` | [Download](https://huggingface.co/EdgeFirst/yolov5-det/resolve/main/onnx/yolov5s-det-coco.onnx) |
92
  | Medium | `yolov5m-det-coco.onnx` | [Download](https://huggingface.co/EdgeFirst/yolov5-det/resolve/main/onnx/yolov5m-det-coco.onnx) |
@@ -96,84 +94,18 @@ Full pipeline timing: pre-processing + inference + post-processing.
96
  </details>
97
 
98
  <details>
99
- <summary><strong>TFLite INT8</strong> — Any platform with TFLite (i.MX 8M Plus uses VX delegate)</summary>
100
-
101
- | Size | File | Download |
102
- |------|------|----------|
103
- | Nano | `yolov5n-det-coco-int8.tflite` | [Download](https://huggingface.co/EdgeFirst/yolov5-det/resolve/main/tflite/yolov5n-det-coco-int8.tflite) |
104
- | Small | `yolov5s-det-coco-int8.tflite` | [Download](https://huggingface.co/EdgeFirst/yolov5-det/resolve/main/tflite/yolov5s-det-coco-int8.tflite) |
105
- | Medium | `yolov5m-det-coco-int8.tflite` | [Download](https://huggingface.co/EdgeFirst/yolov5-det/resolve/main/tflite/yolov5m-det-coco-int8.tflite) |
106
- | Large | `yolov5l-det-coco-int8.tflite` | [Download](https://huggingface.co/EdgeFirst/yolov5-det/resolve/main/tflite/yolov5l-det-coco-int8.tflite) |
107
- | XLarge | `yolov5x-det-coco-int8.tflite` | [Download](https://huggingface.co/EdgeFirst/yolov5-det/resolve/main/tflite/yolov5x-det-coco-int8.tflite) |
108
-
109
- </details>
110
-
111
- ### Platform-Specific
112
-
113
- <details>
114
- <summary><strong>NXP i.MX 93</strong> — Ethos-U NPU via ARM VELA compiler.</summary>
115
 
116
- | Size | File | Download |
117
- |------|------|----------|
118
- | Nano | `yolov5n-det-coco.tflite` | [Download](https://huggingface.co/EdgeFirst/yolov5-det/resolve/main/imx93/yolov5n-det-coco.tflite) |
119
- | Small | `yolov5s-det-coco.tflite` | [Download](https://huggingface.co/EdgeFirst/yolov5-det/resolve/main/imx93/yolov5s-det-coco.tflite) |
120
- | Medium | `yolov5m-det-coco.tflite` | [Download](https://huggingface.co/EdgeFirst/yolov5-det/resolve/main/imx93/yolov5m-det-coco.tflite) |
121
- | Large | `yolov5l-det-coco.tflite` | [Download](https://huggingface.co/EdgeFirst/yolov5-det/resolve/main/imx93/yolov5l-det-coco.tflite) |
122
- | XLarge | `yolov5x-det-coco.tflite` | [Download](https://huggingface.co/EdgeFirst/yolov5-det/resolve/main/imx93/yolov5x-det-coco.tflite) |
123
 
124
  </details>
125
 
126
- <details>
127
- <summary><strong>NXP i.MX 95</strong> — eIQ Neutron NPU optimized.</summary>
128
-
129
- | Size | File | Download |
130
- |------|------|----------|
131
- | Nano | `yolov5n-det-coco.tflite` | [Download](https://huggingface.co/EdgeFirst/yolov5-det/resolve/main/imx95/yolov5n-det-coco.tflite) |
132
- | Small | `yolov5s-det-coco.tflite` | [Download](https://huggingface.co/EdgeFirst/yolov5-det/resolve/main/imx95/yolov5s-det-coco.tflite) |
133
- | Medium | `yolov5m-det-coco.tflite` | [Download](https://huggingface.co/EdgeFirst/yolov5-det/resolve/main/imx95/yolov5m-det-coco.tflite) |
134
- | Large | `yolov5l-det-coco.tflite` | [Download](https://huggingface.co/EdgeFirst/yolov5-det/resolve/main/imx95/yolov5l-det-coco.tflite) |
135
- | XLarge | `yolov5x-det-coco.tflite` | [Download](https://huggingface.co/EdgeFirst/yolov5-det/resolve/main/imx95/yolov5x-det-coco.tflite) |
136
-
137
- </details>
138
-
139
- <details>
140
- <summary><strong>NXP Ara240</strong> — Kinara DVM compiled model.</summary>
141
-
142
- | Size | File | Download |
143
- |------|------|----------|
144
- | Nano | `yolov5n-det-coco.dvm` | [Download](https://huggingface.co/EdgeFirst/yolov5-det/resolve/main/ara240/yolov5n-det-coco.dvm) |
145
- | Small | `yolov5s-det-coco.dvm` | [Download](https://huggingface.co/EdgeFirst/yolov5-det/resolve/main/ara240/yolov5s-det-coco.dvm) |
146
- | Medium | `yolov5m-det-coco.dvm` | [Download](https://huggingface.co/EdgeFirst/yolov5-det/resolve/main/ara240/yolov5m-det-coco.dvm) |
147
- | Large | `yolov5l-det-coco.dvm` | [Download](https://huggingface.co/EdgeFirst/yolov5-det/resolve/main/ara240/yolov5l-det-coco.dvm) |
148
- | XLarge | `yolov5x-det-coco.dvm` | [Download](https://huggingface.co/EdgeFirst/yolov5-det/resolve/main/ara240/yolov5x-det-coco.dvm) |
149
-
150
- </details>
151
-
152
- <details>
153
- <summary><strong>RPi5 + Hailo-8/8L</strong> — Hailo-8L (13 TOPS) and Hailo-8 (26 TOPS).</summary>
154
-
155
- | Size | File | Download |
156
- |------|------|----------|
157
- | Nano | `yolov5n-det-coco.hef` | [Download](https://huggingface.co/EdgeFirst/yolov5-det/resolve/main/hailo/yolov5n-det-coco.hef) |
158
- | Small | `yolov5s-det-coco.hef` | [Download](https://huggingface.co/EdgeFirst/yolov5-det/resolve/main/hailo/yolov5s-det-coco.hef) |
159
- | Medium | `yolov5m-det-coco.hef` | [Download](https://huggingface.co/EdgeFirst/yolov5-det/resolve/main/hailo/yolov5m-det-coco.hef) |
160
- | Large | `yolov5l-det-coco.hef` | [Download](https://huggingface.co/EdgeFirst/yolov5-det/resolve/main/hailo/yolov5l-det-coco.hef) |
161
- | XLarge | `yolov5x-det-coco.hef` | [Download](https://huggingface.co/EdgeFirst/yolov5-det/resolve/main/hailo/yolov5x-det-coco.hef) |
162
-
163
- </details>
164
-
165
- <details>
166
- <summary><strong>NVIDIA Jetson</strong> — Jetson FP16 and INT8 engines.</summary>
167
-
168
- | Size | File | Download |
169
- |------|------|----------|
170
- | Nano | `yolov5n-det-coco.engine` | [Download](https://huggingface.co/EdgeFirst/yolov5-det/resolve/main/jetson/yolov5n-det-coco.engine) |
171
- | Small | `yolov5s-det-coco.engine` | [Download](https://huggingface.co/EdgeFirst/yolov5-det/resolve/main/jetson/yolov5s-det-coco.engine) |
172
- | Medium | `yolov5m-det-coco.engine` | [Download](https://huggingface.co/EdgeFirst/yolov5-det/resolve/main/jetson/yolov5m-det-coco.engine) |
173
- | Large | `yolov5l-det-coco.engine` | [Download](https://huggingface.co/EdgeFirst/yolov5-det/resolve/main/jetson/yolov5l-det-coco.engine) |
174
- | XLarge | `yolov5x-det-coco.engine` | [Download](https://huggingface.co/EdgeFirst/yolov5-det/resolve/main/jetson/yolov5x-det-coco.engine) |
175
-
176
- </details>
177
 
178
 
179
  ---
@@ -189,7 +121,7 @@ gst-launch-1.0 \
189
  v4l2src device=/dev/video0 ! video/x-raw,width=640,height=480 ! \
190
  edgefirstcameraadaptor ! \
191
  tensor_filter framework=tensorflow-lite \
192
- model=yolov5n-det-coco-int8.tflite \
193
  custom=Delegate:External,ExtDelegateLib:libvx_delegate.so ! \
194
  edgefirstdetdecoder ! edgefirstoverlay ! waylandsink
195
  ```
@@ -199,7 +131,7 @@ gst-launch-1.0 \
199
  ```bash
200
  gst-launch-1.0 \
201
  v4l2src device=/dev/video0 ! video/x-raw,width=640,height=480 ! \
202
- hailonet hef-path=yolov5n-det-coco-h8l.hef ! \
203
  hailofilter function-name=yolov5_nms ! \
204
  hailooverlay ! videoconvert ! autovideosink
205
  ```
@@ -225,7 +157,7 @@ gst-launch-1.0 \
225
  from edgefirst.hal import Model, TensorImage
226
 
227
  # Load model — metadata (labels, decoder config) is embedded in the file
228
- model = Model("yolov5n-det-coco-int8.tflite")
229
 
230
  # Run inference on an image
231
  image = TensorImage.from_file("image.jpg")
 
80
 
81
  ## Downloads
82
 
83
+ <details open>
84
+ <summary><strong>ONNX FP32</strong> — Any platform with ONNX Runtime.</summary>
85
 
86
+ | Size | File | Status |
87
+ |------|------|--------|
 
 
 
88
  | Nano | `yolov5n-det-coco.onnx` | [Download](https://huggingface.co/EdgeFirst/yolov5-det/resolve/main/onnx/yolov5n-det-coco.onnx) |
89
  | Small | `yolov5s-det-coco.onnx` | [Download](https://huggingface.co/EdgeFirst/yolov5-det/resolve/main/onnx/yolov5s-det-coco.onnx) |
90
  | Medium | `yolov5m-det-coco.onnx` | [Download](https://huggingface.co/EdgeFirst/yolov5-det/resolve/main/onnx/yolov5m-det-coco.onnx) |
 
94
  </details>
95
 
96
  <details>
97
+ <summary><strong>TFLite INT8</strong> — CPU or NPU via runtime delegate (i.MX 8M Plus VX Delegate).</summary>
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
98
 
99
+ | Size | File | Status |
100
+ |------|------|--------|
101
+ | Nano | `yolov5n-det-coco.tflite` | [Download](https://huggingface.co/EdgeFirst/yolov5-det/resolve/main/tflite/yolov5n-det-coco.tflite) |
102
+ | Small | `yolov5s-det-coco.tflite` | [Download](https://huggingface.co/EdgeFirst/yolov5-det/resolve/main/tflite/yolov5s-det-coco.tflite) |
103
+ | Medium | `yolov5m-det-coco.tflite` | [Download](https://huggingface.co/EdgeFirst/yolov5-det/resolve/main/tflite/yolov5m-det-coco.tflite) |
104
+ | Large | `yolov5l-det-coco.tflite` | [Download](https://huggingface.co/EdgeFirst/yolov5-det/resolve/main/tflite/yolov5l-det-coco.tflite) |
105
+ | XLarge | `yolov5x-det-coco.tflite` | [Download](https://huggingface.co/EdgeFirst/yolov5-det/resolve/main/tflite/yolov5x-det-coco.tflite) |
106
 
107
  </details>
108
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
109
 
110
 
111
  ---
 
121
  v4l2src device=/dev/video0 ! video/x-raw,width=640,height=480 ! \
122
  edgefirstcameraadaptor ! \
123
  tensor_filter framework=tensorflow-lite \
124
+ model=yolov5n-det-coco.tflite \
125
  custom=Delegate:External,ExtDelegateLib:libvx_delegate.so ! \
126
  edgefirstdetdecoder ! edgefirstoverlay ! waylandsink
127
  ```
 
131
  ```bash
132
  gst-launch-1.0 \
133
  v4l2src device=/dev/video0 ! video/x-raw,width=640,height=480 ! \
134
+ hailonet hef-path=yolov5n-det-coco.hailo8l.hef ! \
135
  hailofilter function-name=yolov5_nms ! \
136
  hailooverlay ! videoconvert ! autovideosink
137
  ```
 
157
  from edgefirst.hal import Model, TensorImage
158
 
159
  # Load model — metadata (labels, decoder config) is embedded in the file
160
+ model = Model("yolov5n-det-coco.tflite")
161
 
162
  # Run inference on an image
163
  image = TensorImage.from_file("image.jpg")