Andrey Scherbin
commited on
Commit
·
45419dc
1
Parent(s):
3396e14
Initial commit
Browse files- .gitattributes +1 -0
- README.md +35 -0
- accelerated.gif +3 -0
- baseline.gif +3 -0
- optimized_supercombo_plane.onnx +3 -0
- supercombo_plane.onnx +3 -0
.gitattributes
CHANGED
|
@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
|
|
| 33 |
*.zip filter=lfs diff=lfs merge=lfs -text
|
| 34 |
*.zst filter=lfs diff=lfs merge=lfs -text
|
| 35 |
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
|
|
|
|
|
| 33 |
*.zip filter=lfs diff=lfs merge=lfs -text
|
| 34 |
*.zst filter=lfs diff=lfs merge=lfs -text
|
| 35 |
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
| 36 |
+
*.gif filter=lfs diff=lfs merge=lfs -text
|
README.md
CHANGED
|
@@ -1,3 +1,38 @@
|
|
| 1 |
---
|
| 2 |
license: apache-2.0
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
---
|
| 2 |
license: apache-2.0
|
| 3 |
+
datasets:
|
| 4 |
+
- comma2k19
|
| 5 |
+
tags:
|
| 6 |
+
- onnx
|
| 7 |
+
- ENOT-AutoDL
|
| 8 |
+
- ADAS
|
| 9 |
---
|
| 10 |
+
|
| 11 |
+
# ENOT-AutoDL supercombo optimization on Comma2k19 dataset.
|
| 12 |
+
|
| 13 |
+
This repository contains the modification for
|
| 14 |
+
[supercombo](https://github.com/commaai/openpilot/tree/7d3ad941bc4ba4c923af7a1d7b48544bfc0d3e13/models) model
|
| 15 |
+
with **one target task: plane prediction**.
|
| 16 |
+
Model architecture from [openpilot-pipeline](https://github.com/mbalesni/openpilot-pipeline/blob/main/common/models/supercombo.onnx) repository is used as a baseline.
|
| 17 |
+
Sample videos was generated using [openpilot-pipeline's code](https://github.com/mbalesni/openpilot-pipeline/blob/main/train/vis.ipynb).
|
| 18 |
+
|
| 19 |
+
## Optimization results
|
| 20 |
+
|
| 21 |
+
We use MACs as a latency measure because this metric is device-agnostic and implementation independent.
|
| 22 |
+
There is also a possibility to optimize a model by target device latency using ENOT neural architecture selection algorithm.
|
| 23 |
+
Please, keep in mind that acceleration by device latency differs from acceleration by MACs.
|
| 24 |
+
|
| 25 |
+
| **Model** | **MACs** | **acceleration** |
|
| 26 |
+
|---------------------------------|:---------:|:----------------:|
|
| 27 |
+
| supercombo_plane.onnx | 472774912 | 1.0 |
|
| 28 |
+
| optimized_supercombo_plane.onnx | 109917744 | 4.3 |
|
| 29 |
+
|
| 30 |
+
## Inference example
|
| 31 |
+
|
| 32 |
+
### Baseline model
|
| 33 |
+
<img src="baseline.gif" width="640" height="480"/>
|
| 34 |
+
|
| 35 |
+
### Optimized model
|
| 36 |
+
<img src="accelerated.gif" width="640" height="480"/>
|
| 37 |
+
|
| 38 |
+
If you want to book a demo, please contact us: enot@enot.ai .
|
accelerated.gif
ADDED
|
Git LFS Details
|
baseline.gif
ADDED
|
Git LFS Details
|
optimized_supercombo_plane.onnx
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:608aaef4b4ab69c7a2c33acc21b3d5fa442bb2e882df5ec1b5459332f557949b
|
| 3 |
+
size 24620442
|
supercombo_plane.onnx
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9de8f1681ca475dc55a5d83c145a2a9848abd468f5e80a846f87c2de504f64d7
|
| 3 |
+
size 50524756
|