Skip to content

Commit c3a9168

Browse files
BUG1989BUG1989
authored
update docs, rename ax-pi to axera-pi (#37)
1 parent 7c20ffd commit c3a9168

10 files changed

+149
-22
lines changed

‎README.md

+3-3
Original file line numberDiff line numberDiff line change
@@ -11,13 +11,13 @@
1111

1212
已支持芯片
1313

14-
- AX630A
14+
- [AX630A](docs/AX630A.md)
1515
- [AX620A](docs/AX620A.md)
1616
- [AX620U](docs/AX620U.md)
1717

18-
社区开发板
18+
已支持开发板
1919

20-
- [AX-Pi](https://item.taobao.com/item.htm?_u=m226ocm5e25&id=682169792430)
20+
- [AXera-Pi](https://item.taobao.com/item.htm?_u=m226ocm5e25&id=682169792430)(AX620A)
2121

2222
## 快速上手
2323

‎docs/AX620A.md

+2
Original file line numberDiff line numberDiff line change
@@ -24,3 +24,5 @@
2424
- Ethernet
2525
- One Gigabit Ethernet port
2626
- RGMII and RMII modes
27+
- USB
28+
- USB 2.0 x 1

‎docs/AX620U.md

+2-1
Original file line numberDiff line numberDiff line change
@@ -24,4 +24,5 @@
2424
- Ethernet
2525
- One Gigabit Ethernet port
2626
- RGMII and RMII modes
27-
27+
- USB
28+
- USB 2.0 x 1

‎docs/AX630A.md

+32
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,32 @@
1+
# AX630A
2+
3+
## Brief
4+
5+
- CPU
6+
- Quad-core Cortex A53
7+
- NPU
8+
- 28.8Tops@INT4 or 7.2Tops@INT8
9+
- DSP
10+
- Dual VDSP
11+
- ISP
12+
- 4K@60fps
13+
- AI-ISP
14+
- Video Encoder
15+
- H.264/H.265, 2K@60fps
16+
- Video Decoder
17+
- H.264/H.265, 2k@120fps
18+
- Video Interfaces
19+
- Video Input Interfaces
20+
- 4+4 Lane MIPI CSI
21+
- Video Output Interfaces
22+
- HDMI 2.0
23+
- 4k@60fps
24+
- DDR
25+
- 2 x 32bit LPDDR4x(x)
26+
- Up to 4266Mbit/s rate
27+
- Ethernet
28+
- Two Gigabit Ethernet port
29+
- RGMII and RMII modes
30+
- USB
31+
- USB 3.0 x 2
32+
- USB 2.0 x 1

‎docs/hrnet.png

255 KB
Loading

‎docs/monodlex.png

636 KB
Loading

‎docs/scrfd.jpg

264 KB
Loading

‎docs/yolov5s.jpg

109 KB
Loading

‎docs/yolov7-tiny.jpg

108 KB
Loading

‎examples/README.md

+110-18
Original file line numberDiff line numberDiff line change
@@ -49,7 +49,7 @@ AX-Samples 将不断更新最流行的、实用的、有趣的示例代码。
4949
登录 AX620A 开发板，在 `root` 路径下创建 `samples` 文件夹。
5050

5151
- 将 [快速编译](../docs/compile.md) 中编译生成的可执行示例拷贝到 `/root/samples/` 路径下；
52-
- 将 **[ModelZoo](https://pan.baidu.com/s/1zm2M-vqiss4Rmk-uSoGO7w)** (*pwd: euy7*) 中相应的 **joint** 模型 `mobilenetv2.joint` 、 `yolov5s.joint` 拷贝到 `/root/samples/` 路径下；
52+
- 将 **[ModelZoo](https://pan.baidu.com/s/1ZHW2P6Y3lPf2odmj3fo8hA?pwd=sow9)** 中相应的 **joint** 模型 `mobilenetv2.joint` 、 `yolov5s.joint` 拷贝到 `/root/samples/` 路径下；
5353
- 将测试图片拷贝到 `/root/samples` 路径下。
5454

5555
```
@@ -86,39 +86,64 @@ Tools version: 0.6.0.32
8686
--------------------------------------
8787
Create handle took 201.32 ms (neu 6.36 ms, axe 0.00 ms, overhead 194.96 ms)
8888
--------------------------------------
89-
Repeat 10 times, avg time 4.20 ms, max_time 4.67 ms, min_time 4.14 ms
89+
Repeat 10 times, avg time 3.43 ms, max_time 3.75 ms, min_time 3.37 ms
9090
```
9191

9292
- 物体检测：YOLOv5s
9393

9494
```
95-
/root/samples # ./ax_yolov5s -m yolov5s.joint -i dog.jpg -r 10
95+
root@AXERA:~/samples# ./ax_yolov5s -m yolov5s.joint -i ssd_dog.jpg -r 10
9696
--------------------------------------
97-
model file : yolov5s.joint
98-
image file : dog.jpg
97+
model file : models/yolov5s.joint
98+
image file : images/ssd_dog.jpg
9999
img_h, img_w : 640 640
100-
Run-Joint Runtime version: 0.5.8
100+
Run-Joint Runtime version: 0.5.10
101101
--------------------------------------
102102
[INFO]: Virtual npu mode is 1_1
103103
104-
Tools version: 0.6.0.32
105-
8a011dfa
104+
Tools version: d696ee2f
106105
run over: output len 3
107-
post process cost time:5.52 ms
108106
--------------------------------------
109-
Create handle took 396.61 ms (neu 18.13 ms, axe 0.00 ms, overhead 378.48 ms)
107+
Create handle took 488.94 ms (neu 22.83 ms, axe 0.00 ms, overhead 466.11 ms)
110108
--------------------------------------
111-
Repeat 10 times, avg time 26.59 ms, max_time 27.18 ms, min_time 26.51 ms
109+
Repeat 10 times, avg time 22.54 ms, max_time 22.91 ms, min_time 22.47 ms
112110
--------------------------------------
113111
detection num: 3
114-
16: 92%, [ 133, 219, 312, 543], dog
115-
2: 81%, [ 470, 77, 692, 170], car
116-
1: 60%, [ 169, 120, 565, 417], bicycle
112+
16: 92%, [ 133, 221, 312, 541], dog
113+
2: 77%, [ 468, 76, 692, 171], car
114+
1: 65%, [ 167, 120, 564, 417], bicycle
115+
```
116+
![YOLOv5s](../docs/yolov5s.jpg)
117+
118+
- 物体检测：YOLOv7-Tiny
117119
```
120+
root@AXERA:~/samples# ./ax_yolov7 -m yolov7-tiny.joint -i ssd_dog.jpg -r 10
121+
--------------------------------------
122+
model file : yolov7-tiny-cut-sim-sigmoid-dfs.joint
123+
image file : images/ssd_dog.jpg
124+
img_h, img_w : 416 416
125+
Run-Joint Runtime version: 0.5.10
126+
--------------------------------------
127+
[INFO]: Virtual npu mode is 1_1
128+
129+
Tools version: 0.6.1.4
130+
59588c54
131+
run over: output len 3
132+
--------------------------------------
133+
Create handle took 376.32 ms (neu 15.51 ms, axe 0.00 ms, overhead 360.81 ms)
134+
--------------------------------------
135+
Repeat 10 times, avg time 9.68 ms, max_time 10.01 ms, min_time 9.63 ms
136+
--------------------------------------
137+
detection num: 3
138+
16: 88%, [ 133, 221, 316, 543], dog
139+
1: 86%, [ 139, 130, 571, 422], bicycle
140+
2: 63%, [ 468, 76, 691, 169], car
141+
```
142+
![YOLOv7-Tiny](../docs/yolov7-tiny.jpg)
118143

119144
- 物体检测：YOLOX-S
120145
```
121-
/tmp/qtang # ./ax_yoloxs -m yolox_s_cut.joint -i dog.jpg -r 10
146+
/tmp/samples # ./ax_yoloxs -m yolox_s_cut.joint -i dog.jpg -r 10
122147
--------------------------------------
123148
model file : yolox_s_cut.joint
124149
image file : dog.jpg
@@ -142,6 +167,73 @@ detection num: 4
142167
58: 53%, [ 685, 111, 716, 154], potted plant
143168
```
144169

170+
- 人脸检测：Scrfd
171+
```
172+
root@AXERA:~/samples# ./ax_scrfd -m scrfd_500m_bnkps_shape640x640.joint -i selfie.jpg -r 10
173+
--------------------------------------
174+
model file : models/scrfd_500m_bnkps_shape640x640.joint
175+
image file : images/selfie.jpg
176+
img_h, img_w : 640 640
177+
Run-Joint Runtime version: 0.5.10
178+
--------------------------------------
179+
[INFO]: Virtual npu mode is 1_1
180+
181+
Tools version: 0.6.0.34
182+
9c2b134d
183+
run over: output len 9
184+
--------------------------------------
185+
Create handle took 68.89 ms (neu 4.43 ms, axe 0.00 ms, overhead 64.46 ms)
186+
--------------------------------------
187+
Repeat 10 times, avg time 5.76 ms, max_time 6.09 ms, min_time 5.71 ms
188+
--------------------------------------
189+
detection num: 111
190+
```
191+
![Scrfd](../docs/scrfd.jpg)
192+
193+
- 3D单目车辆检测：Monodlex
194+
```
195+
root@AXERA:~/samples# ./ax_monodlex -m monodlex_sigmoid_max.joint -i cityscape.png -r 10
196+
--------------------------------------
197+
model file : models/monodlex_sigmoid_max.joint
198+
image file : images/cityscape.png
199+
img_h, img_w : 384 1280
200+
Run-Joint Runtime version: 0.5.10
201+
--------------------------------------
202+
[INFO]: Virtual npu mode is 1_1
203+
204+
Tools version: Unknown
205+
run over: output len 8
206+
--------------------------------------
207+
Create handle took 628.84 ms (neu 29.80 ms, axe 0.00 ms, overhead 599.03 ms)
208+
--------------------------------------
209+
Repeat 10 times, avg time 111.77 ms, max_time 112.34 ms, min_time 111.66 ms
210+
--------------------------------------
211+
detection num: 7
212+
```
213+
![Monodlex](../docs/monodlex.png)
214+
215+
- 人体关键点：HRNet
216+
```
217+
root@AXERA:~/samples# ./ax_hrnet -m hrnet_256x192.joint -i pose-1.jpeg -r 10
218+
--------------------------------------
219+
model file : models/hrnet_256x192.joint
220+
image file : images/pose-1.jpeg
221+
img_h, img_w : 256 192
222+
Run-Joint Runtime version: 0.5.10
223+
--------------------------------------
224+
[INFO]: Virtual npu mode is 1_1
225+
226+
Tools version: 0.6.0.30
227+
100b6396
228+
run over: output len 1
229+
--------------------------------------
230+
Create handle took 1385.15 ms (neu 25.64 ms, axe 0.00 ms, overhead 1359.51 ms)
231+
--------------------------------------
232+
Repeat 10 times, avg time 14.11 ms, max_time 14.64 ms, min_time 14.04 ms
233+
--------------------------------------
234+
```
235+
![HRNet](../docs/hrnet.png)
236+
145237
## 模型说明
146238
### YOLOv3(Paddle)
147239
YOLOv3(Paddle) 源自国内产业级深度学习开源框架飞桨的目标检测开发套件 [PaddleDetection](https://github.com/PaddlePaddle/PaddleDetection)，通过速度与精度权衡，我们选择基于 416尺度的 [YOLOv3-Res34](https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/yolov3/) 进行功能展示。
@@ -159,7 +251,7 @@ YOLOv3(Paddle) 源自国内产业级深度学习开源框架飞桨的目标检
159251
#### Sample
160252

161253
```
162-
root@AXERA:~/qtang# ./ax_paddle_yolov3 -m yolov3_paddle.joint -i dog.jpg -r 10
254+
root@AXERA:~/samples# ./ax_paddle_yolov3 -m yolov3_paddle.joint -i dog.jpg -r 10
163255
--------------------------------------
164256
model file : yolov3_paddle.joint
165257
image file : dog.jpg
@@ -198,7 +290,7 @@ MobileSeg 源自国内产业级深度学习开源框架飞桨的高性能图像
198290
#### Sample
199291

200292
```
201-
root@AXERA:~/qtang# ./ax_paddle_mobileseg -m model_mv2seg_sim_cut_infer_argmax.joint -i mv2seg.png -r 10
293+
root@AXERA:~/samples# ./ax_paddle_mobileseg -m model_mv2seg_sim_cut_infer_argmax.joint -i mv2seg.png -r 10
202294
--------------------------------------
203295
model file : model_mv2seg_sim_cut_infer_argmax.joint
204296
image file : mv2seg.png
@@ -231,7 +323,7 @@ PP-HumanSeg 源自国内产业级深度学习开源框架飞桨的高性能图
231323
#### Sample
232324

233325
```
234-
root@AXERA:~/qtang# ./ax_paddle_mobilehumseg -m pp_human_seg_mobile_sim.joint -i pose-1.jpeg -r 10
326+
root@AXERA:~/samples# ./ax_paddle_mobilehumseg -m pp_human_seg_mobile_sim.joint -i pose-1.jpeg -r 10
235327
--------------------------------------
236328
model file : pp_human_seg_mobile_sim.joint
237329
image file : pose-1.jpeg

0 commit comments

Comments
 (0)