Skip to content

P
PaddleDetection

PaddlePaddle / PaddleDetection
1 年多 前同步成功

通知 696
Star 11112
Fork 2696
P
PaddleDetection
未验证 提交 6ceed758 编写于 作者: F Feng Ni 提交者： GitHub
浏览文件
操作

[cherry-pick] update ppyoloe tiny relu model (#7748)

上级 0f4d67a5
隐藏空白更改
内联 并排
Showing with 35 addition and 332 deletion
configs/pphuman/README.md
浏览文件 @ 6ceed758
......@@ -5,14 +5,15 @@
PaddleDetection团队提供了针对行人的基于PP-YOLOE的检测模型，用户可以下载模型进行使用。PP-Human中使用模型为业务数据集模型，我们同时提供CrowdHuman训练配置，可以使用开源数据进行训练。
其中整理后的COCO格式的CrowdHuman数据集[下载链接](https://bj.bcebos.com/v1/paddledet/data/crowdhuman.zip)，检测类别仅一类 `pedestrian(1)`，原始数据集[下载链接](http://www.crowdhuman.org/download.html)。
相关模型的部署模型均在[PP-Human](../../deploy/pipeline/)项目中使用。
| 模型 | 数据集 | mAP<sup>val<br>0.5:0.95 | mAP<sup>val<br>0.5 | 下载 | 配置文件 |
|:---------|:-------:|:------:|:------:| :----: | :------:|
|PP-YOLOE-s| CrowdHuman | 42.5 | 77.9 | [下载链接](https://paddledet.bj.bcebos.com/models/ppyoloe_crn_s_36e_crowdhuman.pdparams) | [配置文件](./ppyoloe_crn_s_36e_crowdhuman.yml) |
|PP-YOLOE-l| CrowdHuman | 48.0 | 81.9 | [下载链接](https://paddledet.bj.bcebos.com/models/ppyoloe_crn_l_36e_crowdhuman.pdparams) | [配置文件](./ppyoloe_crn_l_36e_crowdhuman.yml) |
|PP-YOLOE-s| 业务数据集 | 53.2 | - | [下载链接](https://bj.bcebos.com/v1/paddledet/models/pipeline/mot_ppyoloe_s_36e_pipeline.zip) | [配置文件](./ppyoloe_crn_s_36e_pphuman.yml) |
|PP-YOLOE-l| 业务数据集 | 57.8 | - | [下载链接](https://bj.bcebos.com/v1/paddledet/models/pipeline/mot_ppyoloe_l_36e_pipeline.zip) | [配置文件](./ppyoloe_crn_l_36e_pphuman.yml) |
|PP-YOLOE+_t-P2-aux(320)| 业务数据集 | 49.8 | 85.0 | [下载链接](https://bj.bcebos.com/v1/paddledet/models/pipeline/pphuman/ppyoloe_plus_crn_t_p2_auxhead_320_60e_pphuman.zip) | [配置文件](./ppyoloe_plus_crn_t_p2_auxhead_320_60e_pphuman.yml) |
|PP-YOLOE+_t-aux(320)| 业务数据集 | 45.7 | 81.2 | [下载链接](https://bj.bcebos.com/v1/paddledet/models/pipeline/pphuman/ppyoloe_plus_crn_t_auxhead_320_60e_pphuman.zip) | [配置文件](./ppyoloe_plus_crn_t_auxhead_320_60e_pphuman.yml) |
|PP-YOLOE-s| 业务数据集 | 53.2 | - | [下载链接](https://paddledet.bj.bcebos.com/models/ppyoloe_crn_s_36e_pphuman.pdparams) | [配置文件](./ppyoloe_crn_s_36e_pphuman.yml) |
|PP-YOLOE-l| 业务数据集 | 57.8 | - | [下载链接](https://paddledet.bj.bcebos.com/models/ppyoloe_crn_l_36e_pphuman.pdparams) | [配置文件](./ppyoloe_crn_l_36e_pphuman.yml) |
|PP-YOLOE+_t-aux(320)| 业务数据集 | 45.7 | 81.2 | [下载链接](https://paddledet.bj.bcebos.com/models/pyoloe_plus_crn_t_auxhead_320_60e_pphuman.pdparams) | [配置文件](./ppyoloe_plus_crn_t_auxhead_320_60e_pphuman.yml) |
**注意:**
......
configs/pphuman/ppyoloe_plus_crn_t_p2_auxhead_320_60e_pphuman.yml 已删除 100644 → 0
浏览文件 @ 0f4d67a5
_BASE_: [
'../datasets/coco_detection.yml',
'../runtime.yml',
'../ppyoloe/_base_/optimizer_300e.yml',
'../ppyoloe/_base_/ppyoloe_plus_crn_tiny_auxhead.yml',
'../ppyoloe/_base_/ppyoloe_plus_reader_320.yml',
]
log_iter: 100
snapshot_epoch: 4
weights: output/ppyoloe_plus_crn_t_p2_auxhead_320_60e_pphuman/model_final
pretrain_weights: https://paddledet.bj.bcebos.com/models/ppyoloe_plus_crn_t_p2_auxhead_320_300e_coco.pdparams # 320*320 COCO mAP 36.3
depth_mult: 0.33
width_mult: 0.375
num_classes: 1
TrainDataset:
!COCODataSet
image_dir: ""
anno_path: annotations/train.json
dataset_dir: dataset/pphuman
data_fields: ['image', 'gt_bbox', 'gt_class', 'is_crowd']
EvalDataset:
!COCODataSet
image_dir: ""
anno_path: annotations/val.json
dataset_dir: dataset/pphuman
TestDataset:
!ImageFolder
anno_path: annotations/val.json
dataset_dir: dataset/pphuman
TrainReader:
batch_size: 8
epoch: 60
LearningRate:
base_lr: 0.001
schedulers:
- !CosineDecay
max_epochs: 72
- !LinearWarmup
start_factor: 0.
epochs: 1
architecture: PPYOLOEWithAuxHead
PPYOLOEWithAuxHead:
backbone: CSPResNet
neck: CustomCSPPAN
yolo_head: PPYOLOEHead
aux_head: SimpleConvHead
post_process: ~
CSPResNet:
return_idx: [0, 1, 2, 3] # index 0 stands for P2
CustomCSPPAN:
out_channels: [384, 384, 384, 384]
SimpleConvHead:
fpn_strides: [32, 16, 8, 4]
PPYOLOEHead:
fpn_strides: [32, 16, 8, 4]
static_assigner_epoch: -1
nms:
name: MultiClassNMS
nms_top_k: 1000
keep_top_k: 300
score_threshold: 0.01
nms_threshold: 0.7
configs/ppvehicle/README.md
浏览文件 @ 6ceed758
......@@ -9,6 +9,7 @@ PaddleDetection团队提供了针对自动驾驶场景的基于PP-YOLOE的检测
- PPVehicle9cls数据集整合了BDD100K-MOT和UA-DETRAC，具体类别为9类，包括`pedestrian(1), rider(2), car(3), truck(4), bus(5), van(6), motorcycle(7), bicycle(8), others(9)`。
- PPVehicle数据集整合了BDD100K-MOT和UA-DETRAC，是将BDD100K-MOT中的`car, truck, bus, van`和UA-DETRAC中的`car, bus, van`都合并为1类`vehicle(1)`后的数据集。
相关模型的部署模型均在[PP-Vehicle](../../deploy/pipeline/)项目中使用。
| 模型 | 数据集 | 类别数 | mAP<sup>val<br>0.5:0.95 | 下载链接 | 配置文件 |
|:---------|:---------------:|:------:|:-----------------------:|:---------:| :-----: |
......@@ -19,8 +20,7 @@ PaddleDetection团队提供了针对自动驾驶场景的基于PP-YOLOE的检测
|PP-YOLOE-s| PPVehicle9cls | 9 | 35.3 | [下载链接](https://paddledet.bj.bcebos.com/models/mot_ppyoloe_s_36e_ppvehicle9cls.pdparams) | [配置文件](./mot_ppyoloe_s_36e_ppvehicle9cls.yml) |
|PP-YOLOE-l| PPVehicle | 1 | 63.9 | [下载链接](https://paddledet.bj.bcebos.com/models/mot_ppyoloe_l_36e_ppvehicle.pdparams) | [配置文件](./mot_ppyoloe_l_36e_ppvehicle.yml) |
|PP-YOLOE-s| PPVehicle | 1 | 61.3 | [下载链接](https://paddledet.bj.bcebos.com/models/mot_ppyoloe_s_36e_ppvehicle.pdparams) | [配置文件](./mot_ppyoloe_s_36e_ppvehicle.yml) |
|PP-YOLOE+_t-P2-aux(320)| PPVehicle | 1 | 58.2 | [下载链接](https://bj.bcebos.com/v1/paddledet/models/pipeline/ppvehicle/ppyoloe_plus_crn_t_p2_auxhead_320_60e_ppvehicle.zip) | [配置文件](./ppyoloe_plus_crn_t_p2_auxhead_320_60e_ppvehicle.yml) |
|PP-YOLOE+_t-aux(320)| PPVehicle | 1 | 53.5 | [下载链接](https://bj.bcebos.com/v1/paddledet/models/pipeline/ppvehicle/ppyoloe_plus_crn_t_auxhead_320_60e_ppvehicle.zip) | [配置文件](./ppyoloe_plus_crn_t_auxhead_320_60e_ppvehicle.yml) |
|PP-YOLOE+_t-aux(320)| PPVehicle | 1 | 53.5 | [下载链接](https://paddledet.bj.bcebos.com/models/pipeline/ppyoloe_plus_crn_t_auxhead_320_60e_ppvehicle.pdparams) | [配置文件](./ppyoloe_plus_crn_t_auxhead_320_60e_ppvehicle.yml) |
**注意:**
......
configs/ppvehicle/ppyoloe_plus_crn_t_p2_auxhead_320_60e_ppvehicle.yml 已删除 100644 → 0
浏览文件 @ 0f4d67a5
_BASE_: [
'../datasets/coco_detection.yml',
'../runtime.yml',
'../ppyoloe/_base_/optimizer_300e.yml',
'../ppyoloe/_base_/ppyoloe_plus_crn_tiny_auxhead.yml',
'../ppyoloe/_base_/ppyoloe_plus_reader_320.yml',
]
log_iter: 100
snapshot_epoch: 4
weights: output/ppyoloe_plus_crn_t_p2_auxhead_320_60e_ppvehicle/model_final
pretrain_weights: https://paddledet.bj.bcebos.com/models/ppyoloe_plus_crn_t_p2_auxhead_320_300e_coco.pdparams # 320*320 COCO mAP 36.3
depth_mult: 0.33
width_mult: 0.375
num_classes: 1
TrainDataset:
!COCODataSet
image_dir: ""
anno_path: annotations/train_all.json
dataset_dir: dataset/ppvehicle
data_fields: ['image', 'gt_bbox', 'gt_class', 'is_crowd']
allow_empty: true
EvalDataset:
!COCODataSet
image_dir: ""
anno_path: annotations/val_all.json
dataset_dir: dataset/ppvehicle
TestDataset:
!ImageFolder
anno_path: annotations/val_all.json
dataset_dir: dataset/ppvehicle
TrainReader:
batch_size: 8
epoch: 60
LearningRate:
base_lr: 0.001
schedulers:
- !CosineDecay
max_epochs: 72
- !LinearWarmup
start_factor: 0.
epochs: 1
architecture: PPYOLOEWithAuxHead
PPYOLOEWithAuxHead:
backbone: CSPResNet
neck: CustomCSPPAN
yolo_head: PPYOLOEHead
aux_head: SimpleConvHead
post_process: ~
CSPResNet:
return_idx: [0, 1, 2, 3] # index 0 stands for P2
CustomCSPPAN:
out_channels: [384, 384, 384, 384]
SimpleConvHead:
fpn_strides: [32, 16, 8, 4]
PPYOLOEHead:
fpn_strides: [32, 16, 8, 4]
static_assigner_epoch: -1
nms:
name: MultiClassNMS
nms_top_k: 1000
keep_top_k: 300
score_threshold: 0.01
nms_threshold: 0.7
configs/ppyoloe/README_cn.md
浏览文件 @ 6ceed758
......@@ -48,9 +48,7 @@ PP-YOLOE由以下方法组成
| 模型 | Epoch | GPU个数 | 每GPU图片个数 | 骨干网络 | 输入尺寸 | Box AP<sup>val<br>0.5:0.95 | Box AP<sup>test<br>0.5:0.95 | Params(M) | FLOPs(G) | V100 FP32(FPS) | V100 TensorRT FP16(FPS) | 模型下载 | 配置文件 |
|:---------------:|:-----:|:---------:|:--------:|:----------:|:----------:|:--------------------------:|:---------------------------:|:---------:|:--------:|:---------------:| :---------------------: |:------------------------------------------------------------------------------------:|:-------------------------------------------:|
| PP-YOLOE+_t-aux(640) | 300 | 8 | 8 | cspresnet-t | 640 | 39.7 | 56.4 | 4.85 | 19.15 | - | 344.8 | [model](https://paddledet.bj.bcebos.com/models/ppyoloe_plus_crn_t_auxhead_300e_coco.pdparams) | [config](./ppyoloe_plus_crn_t_auxhead_300e_coco.yml) |
| PP-YOLOE-t-P2(320) | 300 | 8 | 8 | cspresnet-t | 320 | 34.7 | 50.0 | 6.82 | 4.78 | - | - | [model](https://paddledet.bj.bcebos.com/models/ppyoloe_crn_t_p2_320_300e_coco.pdparams) | [config](./ppyoloe_crn_t_p2_320_300e_coco.yml) |
| PP-YOLOE+_t-P2-aux(320) | 300 | 8 | 8 | cspresnet-t | 320 | 36.3 | 51.7 | 6.00 | 15.46 | - | - | [model](https://paddledet.bj.bcebos.com/models/ppyoloe_plus_crn_t_p2_auxhead_320_300e_coco.pdparams) | [config](./ppyoloe_plus_crn_t_p2_auxhead_320_300e_coco.yml) |
| PP-YOLOE+_t-aux(640)-relu | 300 | 8 | 8 | cspresnet-t | 640 | 36.5 | 53.1 | 3.60 | 12.17 | - | 476.2 | [model](https://paddledet.bj.bcebos.com/models/ppyoloe_plus_crn_t_auxhead_relu_300e_coco.pdparams) | [config](./ppyoloe_plus_crn_t_auxhead_relu_300e_coco.yml) |
### 综合指标
| 模型 | Epoch | AP<sup>0.5:0.95 | AP<sup>0.5 | AP<sup>0.75 | AP<sup>small | AP<sup>medium | AP<sup>large | AR<sup>small | AR<sup>medium | AR<sup>large |
......
configs/ppyoloe/ppyoloe_crn_t_p2_320_300e_coco.yml 已删除 100644 → 0
浏览文件 @ 0f4d67a5
_BASE_: [
'../datasets/coco_detection.yml',
'../runtime.yml',
'./_base_/optimizer_300e.yml',
'./_base_/ppyoloe_crn.yml',
'./_base_/ppyoloe_plus_reader_320.yml',
]
log_iter: 100
snapshot_epoch: 10
weights: output/ppyoloe_crn_t_p2_320_300e_coco/model_final
pretrain_weights: https://paddledet.bj.bcebos.com/models/pretrained/CSPResNetb_t_pretrained.pdparams
depth_mult: 0.33
width_mult: 0.375
CSPResNet:
return_idx: [0, 1, 2, 3]
CustomCSPPAN:
out_channels: [768, 384, 192, 96]
PPYOLOEHead:
fpn_strides: [32, 16, 8, 4]
attn_conv: 'repvgg' #
assigner:
name: TaskAlignedAssigner
topk: 13
alpha: 1.0
beta: 6.0
is_close_gt: True #
nms:
name: MultiClassNMS
nms_top_k: 1000
keep_top_k: 300
score_threshold: 0.01
nms_threshold: 0.7
worker_num: 4
eval_height: &eval_height 320
eval_width: &eval_width 320
eval_size: &eval_size [*eval_height, *eval_width]
TrainReader:
sample_transforms:
- Decode: {}
- RandomDistort: {}
- RandomExpand: {fill_value: [123.675, 116.28, 103.53]}
- RandomCrop: {}
- RandomFlip: {}
batch_transforms:
- BatchRandomResize: {target_size: [224, 256, 288, 320, 352, 384, 416, 448, 480, 512, 544], random_size: True, random_interp: True, keep_ratio: False}
- NormalizeImage: {mean: [0.485, 0.456, 0.406], std: [0.229, 0.224, 0.225], is_scale: True}
- Permute: {}
- PadGT: {}
batch_size: 8
shuffle: true
drop_last: true
use_shared_memory: true
collate_batch: true
EvalReader:
sample_transforms:
- Decode: {}
- Resize: {target_size: *eval_size, keep_ratio: False, interp: 2}
- NormalizeImage: {mean: [0.485, 0.456, 0.406], std: [0.229, 0.224, 0.225], is_scale: True}
- Permute: {}
batch_size: 2
TestReader:
inputs_def:
image_shape: [3, *eval_height, *eval_width]
sample_transforms:
- Decode: {}
- Resize: {target_size: *eval_size, keep_ratio: False, interp: 2}
- NormalizeImage: {mean: [0.485, 0.456, 0.406], std: [0.229, 0.224, 0.225], is_scale: True}
- Permute: {}
batch_size: 1
fuse_normalize: True
configs/ppyoloe/ppyoloe_plus_crn_t_auxhead_300e_coco.yml
浏览文件 @ 6ceed758
......@@ -3,7 +3,7 @@ _BASE_: [
'../runtime.yml',
'./_base_/optimizer_300e.yml',
'./_base_/ppyoloe_plus_crn_tiny_auxhead.yml',
'./_base_/ppyoloe_plus_reader.yml', # 640*640
'./_base_/ppyoloe_plus_reader.yml',
]
log_iter: 100
......@@ -13,27 +13,3 @@ weights: output/ppyoloe_plus_crn_t_auxhead_300e_coco/model_final
pretrain_weights: https://paddledet.bj.bcebos.com/models/pretrained/CSPResNetb_t_pretrained.pdparams
depth_mult: 0.33
width_mult: 0.375
worker_num: 4
eval_height: &eval_height 640
eval_width: &eval_width 640
eval_size: &eval_size [*eval_height, *eval_width]
EvalReader:
sample_transforms:
- Decode: {}
- Resize: {target_size: *eval_size, keep_ratio: False, interp: 2}
- NormalizeImage: {mean: [0., 0., 0.], std: [1., 1., 1.], norm_type: none}
- Permute: {}
batch_size: 2
TestReader:
inputs_def:
image_shape: [3, *eval_height, *eval_width]
sample_transforms:
- Decode: {}
- Resize: {target_size: *eval_size, keep_ratio: False, interp: 2}
- NormalizeImage: {mean: [0., 0., 0.], std: [1., 1., 1.], norm_type: none}
- Permute: {}
batch_size: 1
configs/ppyoloe/ppyoloe_plus_crn_t_auxhead_relu_300e_coco.yml 0 → 100644
浏览文件 @ 6ceed758
_BASE_: [
'../datasets/coco_detection.yml',
'../runtime.yml',
'./_base_/optimizer_300e.yml',
'./_base_/ppyoloe_plus_crn_tiny_auxhead.yml',
'./_base_/ppyoloe_plus_reader.yml',
]
log_iter: 100
snapshot_epoch: 10
weights: output/ppyoloe_plus_crn_t_auxhead_relu_300e_coco/model_final
pretrain_weights: https://paddledet.bj.bcebos.com/models/pretrained/CSPResNetb_t_pretrained.pdparams
depth_mult: 0.33
width_mult: 0.375
CSPResNet:
act: 'relu'
CustomCSPPAN:
act: 'relu'
PPYOLOEHead:
act: 'relu'
attn_conv: None
configs/ppyoloe/ppyoloe_plus_crn_t_p2_auxhead_320_300e_coco.yml 已删除 100644 → 0
浏览文件 @ 0f4d67a5
_BASE_: [
'../datasets/coco_detection.yml',
'../runtime.yml',
'./_base_/optimizer_300e.yml',
'./_base_/ppyoloe_plus_crn_tiny_auxhead.yml',
'./_base_/ppyoloe_plus_reader_320.yml',
]
log_iter: 100
snapshot_epoch: 10
weights: output/ppyoloe_plus_crn_t_p2_auxhead_320_300e_coco/model_final
pretrain_weights: https://paddledet.bj.bcebos.com/models/pretrained/CSPResNetb_t_pretrained.pdparams
depth_mult: 0.33
width_mult: 0.375
architecture: PPYOLOEWithAuxHead
PPYOLOEWithAuxHead:
backbone: CSPResNet
neck: CustomCSPPAN
yolo_head: PPYOLOEHead
aux_head: SimpleConvHead
post_process: ~
CSPResNet:
return_idx: [0, 1, 2, 3] # index 0 stands for P2
CustomCSPPAN:
out_channels: [384, 384, 384, 384]
SimpleConvHead:
fpn_strides: [32, 16, 8, 4]
PPYOLOEHead:
fpn_strides: [32, 16, 8, 4]
worker_num: 4
eval_height: &eval_height 320
eval_width: &eval_width 320
eval_size: &eval_size [*eval_height, *eval_width]
EvalReader:
sample_transforms:
- Decode: {}
- Resize: {target_size: *eval_size, keep_ratio: False, interp: 2}
- NormalizeImage: {mean: [0., 0., 0.], std: [1., 1., 1.], norm_type: none}
- Permute: {}
batch_size: 2
TestReader:
inputs_def:
image_shape: [3, *eval_height, *eval_width]
sample_transforms:
- Decode: {}
- Resize: {target_size: *eval_size, keep_ratio: False, interp: 2}
- NormalizeImage: {mean: [0., 0., 0.], std: [1., 1., 1.], norm_type: none}
- Permute: {}
batch_size: 1
Markdown is supported
0% 或 .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑！
想要评论请 注册 或