o
Ďi8 @ s d dl mZ d dlZd dlmZmZ d dlmZ d dlm Z m
Z
d dl m Z m
Z
d dlmZ d d
efd d ZG d
d dZ d ddZ d!ddZ d"ddZd#ddZ d$ddZdd Zd d ZdS )% )PathN) AutoShapeDetectMultiBackend) attempt_load)ClassificationModelSegmentationModel)LOGGERlogging)
select_deviceTFhf_tokenc
C s |st tj t|}zKt| |||d}|rV|jr&t|jt r&t
d n0|jr5t|jt r5t
d n!zt |}W n t
yU } zt
d| W Y d}~nd}~ww W n! t
yx } zt
d| t| |dd}W Y d}~nd}~ww |st tj ||S ) a
Creates a specified YOLOv5 model
Arguments:
model_path (str): path of the model
device (str): select device that model will be loaded (cpu, cuda)
pretrained (bool): load pretrained weights into the model
autoshape (bool): make model ready for inference
verbose (bool): if False, yolov5 logs will be silent
hf_token (str): huggingface read token for private models
Returns:
pytorch model
(Adapted from yolov5.hubconf.create)
)devicefuser u WARNING ⚠️ YOLOv5 ClassificationModel is not yet AutoShape compatible. You must pass torch tensors in BCHW to this model, i.e. shape(1,3,224,224).u WARNING ⚠️ YOLOv5 SegmentationModel is not yet AutoShape compatible. You will not be able to run inference with this model.u! WARNING ⚠️ autoshape failed: Nu* WARNING ⚠️ DetectMultiBackend failed: F)r r
)r setLevelr WARNINGr
r pt
isinstancemodelr warningr r Exceptionr INFOto)
model_pathr autoshapeverboser r e r J/home/jeff/fluffinator/venv/lib/python3.10/site-packages/yolov5/helpers.py
load_model s>
r c @ s( e Zd Zd ddZdd Zd d d
ZdS )
YOLOv5NTc C sR t dt || _|| _|r$t|jd jddd t||dd| _ d S d | _ d S )NzfYOLOv5 class is deprecated and will be removed in future release. Use 'model = yolov5.load()' instead.r Tparentsexist_okr r r )
warningswarnDeprecationWarningr r r r mkdirr r )selfr r load_on_initr r r __init__D s
zYOLOv5.__init__c C s2 t | jjd jddd t| j| jdd| _dS )z%
Load yolov5 weight.
r Tr r" N)r r r r& r r r )r' r r r r Q s
zYOLOv5.load_model Fc C s& | j dus J d| j |||d}|S )z
Perform yolov5 prediction using loaded model weights.
Returns results as a yolov5.models.common.Detections object.
Nz.before predict, you need to call .load_model())imssizeaugment)r )r'
image_listr, r- resultsr r r predictZ s zYOLOv5.predict)NT)r* F)__name__
__module__ __qualname__r) r r0 r r r r r C s
r object-detectionr* c C s ddl m} |d urd| d}d| d| d}nd }}d | d
| d | d | d
| d| d| d| d| d| d| d| d| dS )Nr ) __version__z
datasets:
-
z
dataset:
type: z
name: z
split: validation
z'
---
tags:
- yolov5
- yolo
- vision
- z1
- pytorch
library_name: yolov5
library_version: z
inference: false
z
model-index:
- name: z"
results:
- task:
type: zh
metrics:
- type: precision # since mAP@0.5 is not available on hf.co/metrics
value: z` # min: 0.0 - max: 1.0
name: mAP@0.5
---
### How to use
- Install [yolov5](https://github.com/fcakyon/yolov5-pip):
```bash
pip install -U yolov5
```
- Load model and perform prediction:
```python
import yolov5
# load model
model = yolov5.load('a ')
# set model parameters
model.conf = 0.25 # NMS confidence threshold
model.iou = 0.45 # NMS IoU threshold
model.agnostic = False # NMS class-agnostic
model.multi_label = False # NMS multiple labels per box
model.max_det = 1000 # maximum number of detections per image
# set image
img = 'https://github.com/ultralytics/yolov5/raw/master/data/images/zidane.jpg'
# perform inference
results = model(img, size=a )
# inference with test time augmentation
results = model(img, augment=True)
# parse results
predictions = results.pred[0]
boxes = predictions[:, :4] # x1, y1, x2, y2
scores = predictions[:, 4]
categories = predictions[:, 5]
# show detection bounding boxes on image
results.show()
# save results into "results/" folder
results.save(save_dir='results/')
```
- Finetune the model on your custom dataset:
```bash
yolov5 train --data data.yaml --img z --batch 16 --weights z} --epochs 10
```
**More models available at: [awesome-yolov5-models](https://github.com/keremberke/awesome-yolov5-models)**
)yolov5r5 )repo_idap50task
input_size
dataset_idyolov5_versiondatasets_str_1datasets_str_2r r r generate_model_usage_markdowne sL
*7LLrA c C s ddl m}m} | | ||dd t|d }
|| t|
dd|dd t| ||||d
} t|d } t| d
}
|
| W d n1 sFw Y || t| t| jd
|dd d S )Nr upload_file create_repoTr9 tokenprivater! zval_batch0_labels.jpgzsample_visuals.jpgzupload sample visualsr r9 path_or_fileobj path_in_repocommit_messagerF repo_type)r; r< r= r: z README.mdwzAdd yolov5 model card) huggingface_hubrC rD r strrA openwritename)r9
exp_folderr: r r< r; rG r= rC rD sample_visual_pathmodelcard_markdownmodelcard_path file_objectr r r push_model_card_to_hfhub sF
rX c
C s ddl m}m} ddl} |||d}
t|d } t| d} | |
| W d n1 s.w Y || ||dd || t| t| jd |d
d dS ) a
Pushes a yolov5 config to huggingface hub
Arguments:
repo_id (str): The name of the repository to create on huggingface.co
exp_folder (str): The path to the experiment folder
best_ap50 (float): The best ap50 score of the model
input_size (int): The input size of the model (default: 640)
task (str): The task of the model (default: object-detection)
hf_token (str): The huggingface token to use to push the model
private (bool): Whether the model should be private or not
r rB N)r< r; best_ap50z config.jsonrM TrE zAdd yolov5 configr rH ) rN rC rD jsonr rP dumprO rR )
r9 rS rY r< r; r rG rC rD rZ config config_pathrW r r r push_config_to_hfhub s*
r^ c C s ddl m}m} ddlm} t|d }|| dd }|| ||dd || t|t|jd |d
d || t|t|jd |d
d d
S )a'
Pushes a yolov5 model to huggingface hub
Arguments:
repo_id (str): huggingface repo id to be uploaded to
exp_folder (str): yolov5 experiment folder path
hf_token (str): huggingface write token
private (bool): whether to make the repo private or not
r rB )globzweights/best.ptz/events.out.tfevents*TrE zUpload yolov5 best modelr rH z Upload yolov5 tensorboard logsN)rN rC rD r_ r rO rR ) r9 rS r rG rC rD r_ best_model_pathtensorboard_log_pathr r r push_model_to_hfhub* s4
rc c
C sz ddl m} ddlm} m}
m} t|d d| | | ||||||d |
| |||||||d | | |||d d S ) Nr colorstr)r^ rX rc zhub:z Pushing to hf.co/)r9 rS rY r< r; r rG )r9 rS r< r; r rG r= r: )r9 rS r rG )yolov5.utils.generalre yolov5.helpersr^ rX rc r info) hf_model_idr
hf_privatesave_dir
hf_dataset_idr< rY r; re r^ rX rc r r r
push_to_hfhubR s2
rm c C s* dd l }ddlm} d}d}t| jdd }||}|dd ur$d}|dd ur-d}W d n1 s7w Y |rE| jrEtd |rdd
l m
} dd l m } d}
| r[d}
|| j|d |
d
} | | _t| dd?}||} |d | d< |d | d< |d | d< |d | d< |dd ur|d | d< |dd ur|d | d< W d n1 sw Y t| d}|
| | W d n1 sw Y |d d }
|
jddd d|v rt|d r||d t|
d d|v rt|d r||d t|
d d S d S d S d S )Nr )copyfileFignoreerrorstrain_json_pathTyolo_s3_data_dirzg'--upload_dataset' argument cannot be passed when 'yolo_s3_data_dir' field is not empty in 'data.yaml'.) export_coco_as_yolov5_via_yml)is_colabdata)yml_path
output_dirdisable_symlink
val_json_pathtrain_image_dir
val_image_dircoco_s3_data_dirrM cocor z
train.jsonzval.json)yamlshutilrn rP rv safe_loadgetupload_dataset
ValueErrorsahi.utils.cocort rf ru r[ r& r is_filerO )optrk r rn is_coco_datahas_yolo_s3_data_dirf data_infort ru ry rv updated_data_inforM r r r convert_coco_dataset_to_yolo} sf
$r c C s dd l }dd l}ddlm} ddlm}m} t|dd
}||} W d n1 s,w Y dt t
| j dd|j
d |jd }
t|d
d |
|
d } || j| d
} |
d }
|| d |
d} |
d }|| d |d} | rt|d
d|
d S d S )Nr rd )upload_file_to_s3upload_folder_to_s3ro rp zs3://r7 rv /zaws:z% Uploading yolo formatted dataset to z
/data.yaml)
local_files3_filez/train/train) local_folder s3_folderz/val/valz+ Dataset has been successfully uploaded to )r osrf re yolov5.utils.awsr r rP r rO r
s3_upload_dirreplacerR sepr rh rv )r rv rk r r re r r r r r r results3_folder_train
s3_folder_valr r r upload_to_s3 s4
r )NTFN)r4 r* N)Nr* r4 FN)Nr* r4 NF)NF)Nr* Nr4 )pathlibr r# yolov5.models.commonr r yolov5.models.experimentalr yolov5.models.yolor r rf r r yolov5.utils.torch_utilsr
rO r r rA rX r^ rc rm r r r r r r sD
8#
i
6
--
+ :