thebears/deepstream_yolo
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Add PP-YOLOE+ support

Browse Source
This commit is contained in:
Marcos Luciano
2023-01-31 02:59:56 -03:00
parent 825d6bfda8
commit 69f29f8934
4 changed files with 61 additions and 24 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
README.md
View File

@@ -7,7 +7,6 @@ NVIDIA DeepStream SDK 6.1.1 / 6.1 / 6.0.1 / 6.0 configuration for YOLO models
* DeepStream tutorials
* YOLOv6 support
* Dynamic batch-size
* PP-YOLOE+ support
### Improvements on this repository
@@ -29,6 +28,7 @@ NVIDIA DeepStream SDK 6.1.1 / 6.1 / 6.0.1 / 6.0 configuration for YOLO models
* Models benchmarks
* **YOLOv8 support**
* **YOLOX support**
* **PP-YOLOE+ support**
##
@@ -44,7 +44,7 @@ NVIDIA DeepStream SDK 6.1.1 / 6.1 / 6.0.1 / 6.0 configuration for YOLO models
* [INT8 calibration](#int8-calibration)
* [YOLOv5 usage](docs/YOLOv5.md)
* [YOLOR usage](docs/YOLOR.md)
* [PP-YOLOE usage](docs/PPYOLOE.md)
* [PP-YOLOE / PP-YOLOE+ usage](docs/PPYOLOE.md)
* [YOLOv7 usage](docs/YOLOv7.md)
* [YOLOv8 usage](docs/YOLOv8.md)
* [YOLOX usage](docs/YOLOX.md)
@@ -110,7 +110,7 @@ NVIDIA DeepStream SDK 6.1.1 / 6.1 / 6.0.1 / 6.0 configuration for YOLO models
* [Darknet YOLO](https://github.com/AlexeyAB/darknet)
* [YOLOv5 >= 2.0](https://github.com/ultralytics/yolov5)
* [YOLOR](https://github.com/WongKinYiu/yolor)
* [PP-YOLOE](https://github.com/PaddlePaddle/PaddleDetection/tree/release/2.4/configs/ppyoloe)
* [PP-YOLOE / PP-YOLOE+](https://github.com/PaddlePaddle/PaddleDetection/tree/release/2.5/configs/ppyoloe)
* [YOLOv7](https://github.com/WongKinYiu/yolov7)
* [YOLOv8](https://github.com/ultralytics/ultralytics)
* [YOLOX](https://github.com/Megvii-BaseDetection/YOLOX)

26
config_infer_primary_ppyoloe_plus.txt Normal file
View File

@@ -0,0 +1,26 @@
[property]
gpu-id=0
net-scale-factor=0.0039215697906911373
model-color-format=0
custom-network-config=ppyoloe_plus_crn_s_80e_coco.cfg
model-file=ppyoloe_plus_crn_s_80e_coco.wts
model-engine-file=model_b1_gpu0_fp32.engine
#int8-calib-file=calib.table
labelfile-path=labels.txt
batch-size=1
network-mode=0
num-detected-classes=80
interval=0
gie-unique-id=1
process-mode=1
network-type=0
cluster-mode=2
maintain-aspect-ratio=0
parse-bbox-func-name=NvDsInferParseYolo
custom-lib-path=nvdsinfer_custom_impl_Yolo/libnvdsinfer_custom_impl_Yolo.so
engine-create-func-name=NvDsInferYoloCudaEngineGet
[class-attrs-all]
nms-iou-threshold=0.7
pre-cluster-threshold=0.25
topk=300

36
docs/PPYOLOE.md
View File

@@ -1,8 +1,8 @@
# PP-YOLOE usage
# PP-YOLOE / PP-YOLOE+ usage
* [Convert model](#convert-model)
* [Compile the lib](#compile-the-lib)
* [Edit the config_infer_primary_ppyoloe file](#edit-the-config_infer_primary_ppyoloe-file)
* [Edit the config_infer_primary_ppyoloe_plus file](#edit-the-config_infer_primary_ppyoloe_plus-file)
* [Edit the deepstream_app_config file](#edit-the-deepstream_app_config-file)
* [Testing the model](#testing-the-model)
@@ -12,7 +12,7 @@
#### 1. Download the PaddleDetection repo and install the requirements
https://github.com/PaddlePaddle/PaddleDetection/blob/release/2.4/docs/tutorials/INSTALL.md
https://github.com/PaddlePaddle/PaddleDetection/blob/release/2.5/docs/tutorials/INSTALL.md
**NOTE**: It is recommended to use Python virtualenv.
@@ -22,20 +22,20 @@ Copy the `gen_wts_ppyoloe.py` file from `DeepStream-Yolo/utils` directory to the
#### 3. Download the model
Download the `pdparams` file from [PP-YOLOE](https://github.com/PaddlePaddle/PaddleDetection/tree/release/2.4/configs/ppyoloe) releases (example for PP-YOLOE-s)
Download the `pdparams` file from [PP-YOLOE](https://github.com/PaddlePaddle/PaddleDetection/tree/release/2.5/configs/ppyoloe) releases (example for PP-YOLOE+_s)
```
wget https://paddledet.bj.bcebos.com/models/ppyoloe_crn_s_400e_coco.pdparams
wget https://paddledet.bj.bcebos.com/models/ppyoloe_plus_crn_s_80e_coco.pdparams
```
**NOTE**: You can use your custom model, but it is important to keep the YOLO model reference (`ppyoloe_`) in you `cfg` and `weights`/`wts` filenames to generate the engine correctly.
#### 4. Convert model
Generate the `cfg` and `wts` files (example for PP-YOLOE-s)
Generate the `cfg` and `wts` files (example for PP-YOLOE+_s)
```
python3 gen_wts_ppyoloe.py -w ppyoloe_crn_s_400e_coco.pdparams -c configs/ppyoloe/ppyoloe_crn_s_400e_coco.yml
python3 gen_wts_ppyoloe.py -w ppyoloe_plus_crn_s_80e_coco.pdparams -c configs/ppyoloe/ppyoloe_plus_crn_s_80e_coco.yml
```
#### 5. Copy generated files
@@ -80,19 +80,27 @@ Open the `DeepStream-Yolo` folder and compile the lib
##
### Edit the config_infer_primary_ppyoloe file
### Edit the config_infer_primary_ppyoloe_plus file
Edit the `config_infer_primary_ppyoloe.txt` file according to your model (example for PP-YOLOE-s)
Edit the `config_infer_primary_ppyoloe_plus.txt` file according to your model (example for PP-YOLOE+_s)
```
[property]
...
custom-network-config=ppyoloe_crn_s_400e_coco.cfg
model-file=ppyoloe_crn_s_400e_coco.wts
custom-network-config=ppyoloe_plus_crn_s_80e_coco.cfg
model-file=ppyoloe_plus_crn_s_80e_coco.wts
...
```
**NOTE**: The PP-YOLOE uses normalization on the image preprocess. It is important to change the `net-scale-factor` and `offsets` according to the trained values.
**NOTE**: If you use the **legacy** model, you should edit the `config_infer_primary_ppyoloe.txt` file.
**NOTE**: The **PP-YOLOE+** uses zero mean normalization on the image preprocess. It is important to change the `net-scale-factor` according to the trained values.
```
net-scale-factor=0.0039215697906911373
```
**NOTE**: The **PP-YOLOE (legacy)** uses normalization on the image preprocess. It is important to change the `net-scale-factor` and `offsets` according to the trained values.
Default: `mean = 0.485, 0.456, 0.406` and `std = 0.229, 0.224, 0.225`
@@ -109,9 +117,11 @@ offsets=123.675;116.28;103.53
...
[primary-gie]
...
config-file=config_infer_primary_ppyoloe.txt
config-file=config_infer_primary_ppyoloe_plus.txt
```
**NOTE**: If you use the **legacy** model, you should edit it to `config_infer_primary_ppyoloe.txt`.
##
### Testing the model

17
utils/gen_wts_ppyoloe.py
View File

@@ -3,7 +3,7 @@ import struct
import paddle
import numpy as np
from ppdet.core.workspace import load_config, merge_config
from ppdet.utils.check import check_gpu, check_version, check_config
from ppdet.utils.check import check_version, check_config
from ppdet.utils.cli import ArgsParser
from ppdet.engine import Trainer
from ppdet.slim import build_slim_model
@@ -273,13 +273,14 @@ class Layers(object):
def get_state_dict(self, state_dict):
for k, v in state_dict.items():
vr = v.reshape([-1]).numpy()
self.fw.write('{} {} '.format(k, len(vr)))
for vv in vr:
self.fw.write(' ')
self.fw.write(struct.pack('>f', float(vv)).hex())
self.fw.write('\n')
self.wc += 1
if 'alpha' not in k:
vr = v.reshape([-1]).numpy()
self.fw.write('{} {} '.format(k, len(vr)))
for vv in vr:
self.fw.write(' ')
self.fw.write(struct.pack('>f', float(vv)).hex())
self.fw.write('\n')
self.wc += 1
def get_anchors(self, anchor_points, stride_tensor):
vr = anchor_points.numpy()