Datasets
Models
Downloads: 1
99 lines (98 with data), 12.5 kB
U
hĤaUE ã @ sÌ d Z ddlZddlZddlZddlZddlmZ ddlmZ ddl Z
ddl Z ddl m Z ee
¡ Zejd Zeeejkrej ee¡ eej ee ¡ ¡ZddlmZ ddlmZ ddlm Z dd l m Z m Z m Z m!Z!m"Z"m#Z#m$Z$m%Z%m&Z&m'Z'm(Z(m)Z)m*Z*m+Z+ dd
l,m-Z-m.Z. dd l/m0Z0m1Z1m2Z2 dd l3m4Z4m5Z5 d
d Z6dd Z7dd Z8e 9¡ dddddddddddddded ddd dddedd e dfd d Z:d d Z;d!d" Z<e=d#krÈe; Z>e<e> dS )$z
Validate a trained YOLOv5 model accuracy on a custom dataset
Usage:
$ python path/to/val.py --data coco128.yaml --weights yolov5s.pt --img 640
é N)ÚPath)ÚThread)Útqdm)ÚDetectMultiBackend)Ú Callbacks)Úcreate_dataloader)ÚLOGGERÚbox_iouÚ
check_datasetÚcheck_img_sizeÚcheck_requirementsÚ
check_yamlÚcoco80_to_coco91_classÚcolorstrÚincrement_pathÚnon_max_suppressionÚ
print_argsÚ scale_coordsÚ xywh2xyxyÚ xyxy2xywh)ÚConfusionMatrixÚ ap_per_class)Úoutput_to_targetÚ plot_imagesÚplot_val_study)Ú
select_deviceÚ time_syncc
C s¤ t |¡ddddg }| ¡ D ]^ }}}tt |¡ dd¡| d¡ ¡ }|r`|f||fn|f|} t|d$}
|
dt| ¡ | d ¡ W 5 Q R X q d S )Né r é éÿÿÿÿÚaz%g Ú
) ÚtorchÚtensorÚtolistr ÚviewÚopenÚwriteÚlenÚrstrip) ÚprednÚ save_confÚshapeÚfileZgnZxyxyZconfÚclsZxywhÚlineÚf© r1 ú1C:\Users\safak\Desktop\blood_cancer\yolov5\val.pyÚ save_one_txt% s $ r3 c C s¶ |j ¡ rt|j n|j }t| d d
d d
f }|d d
d d
f |d d
dd
f d 8 < t| ¡ | ¡ D ]<\}}| ||t|d dd |D t|d dd¡ qtd S )Nr é é c S s g | ]}t |d qS )é )Úround©Ú.0Úxr1 r1 r2 Ú
<listcomp>7 s z!save_one_json.<locals>.<listcomp>)Úimage_idZ category_idÚbboxÚscore)ÚstemÚ isnumericÚintr Úzipr$ Úappendr7 )r* ÚjdictÚpathÚ class_mapr< ÚboxÚpÚbr1 r1 r2 Ú
save_one_json/ s 0 ýrJ c C s t j| jd |jd t j|jd}t|dd
dd
f | dd
dd
f }t ||d k|dd
dd
f | dd
df k@ ¡}|d jd rt t |d¡||d |d f dd
df fd¡ ¡
¡ }|d jd dkrH||dd
df ¡ ddd
}|t j
|dd
df d d
d }|t j
|dd
df d d
d }t |¡ |j¡}|dd
dd
f |k||dd
df ¡ < |S ) a
Return correct predictions matrix. Both sets of boxes are in (x1, y1, x2, y2) format.
Arguments:
detections (Array[N, 6]), x1, y1, x2, y2, conf, class
labels (Array[M, 5]), class, x1, y1, x2, y2
Returns:
correct (Array[N, 10]), for 10 IoU levels
r )ÚdtypeÚdeviceNr r r5 r4 r T)Ú return_indexr6 )r" Úzerosr, ÚboolrL r ÚwhereÚcatÚstackÚcpuÚnumpyÚargsortÚnpÚuniqueÚTensorÚtoÚlong)Z
detectionsÚlabelsÚiouvÚcorrectZiour: Úmatchesr1 r1 r2 Ú
process_batch; s "*6<""",r_ é é çü©ñÒMbP?ç333333ã?ÚvalÚ Fúruns/valÚexpTc ` sF
|d k }|rJt | ¡ jdd }}} ||jdkM }|r@| ¡ n| ¡ nÜt||d}tt|| |d}| rv|d n|j ddd t
|||d}|j |j |j
} }} t|| d }||s¾| oÆ|jdkM }|rè|rÜ|j ¡ n|j ¡ n6| rô|j}n*d}d
}t d¡}t d | d | d
¡ t| } | ¡ t| d¡toL| d d¡} |rXd
n
t| d } t ddd¡ |¡} | ¡ }!|sê|j d
d||f|d |dkrªdnd}"|dkr¼|nd}t | | ||| ||"|t | ddd }d t!| d }#d d t"t#|d r|j$n|j%j$D }$| r,t& n
t't(d }%d!d" }&dddgdddddddf\}'}(})}*}+},}-}.tj)d|d#}/g g g g f\}0}1}2}3t*||&d$d%}4t"|4D ]ö\}5\}6}7}8t+ }9|sÆ| rÞ|6j |dd&}6 |¡|rì|6 ¡ n|6 ¡ }6|6d' }6|6j,\}:};}<}=t+ }>|'d |>|9 7 < |r2||6n ||6| dd(\}?}@|'d
t+ |> 7 < |r||/|d)d* |@D d
7 }/d d
d+d
f t -|=|<|=|<g¡ |¡9 < | rÈfd,d*t(|:D ng }At+ }Bt.|?|||Ad|d-}?|'d+ t+ |B 7 < t"|?D ]$\}C}Dd d
df |Ckd
d
f }Et/|E}F|FrP|Ed d
df 0¡ ng }Gt|7|C |8|C d }H}I d
7 t/|Ddkr¶|Fr|1 1tj)d|!tj2d.t -¡ t -¡ |Gf¡ q|rÌd|Dd d
d/f< |D 3¡ }Jt4|6|C j,d
d
|Jd d
d d0
f |I|8|C d
|Frt5|Ed d
d
d/
f }Kt4|6|C j,d
d
|K|I|8|C d
t 6|Ed d
dd
f |Kfd
¡}Lt7|J|L| }M|r¢|# 7|J|L¡ ntj)|Dj,d |!tj2d.}M|1 1|M 8¡ |Dd d
d0f 8¡ |Dd d
d/f 8¡ |Gf¡ | rüt9|J|
|I|d |Hj:d1 d2 |rt;|J|0|H|% | <d3|D|J|H|$|6|C ¡ q|r¢|5dk r¢|d4|5 d5 }Nt=t>|6|7|N|$fdd6 ?¡ |d4|5 d7 }Nt=t>|6t@|?|7|N|$fdd6 ?¡ q¢d8d* tA|1 D }1t/|1rB|1d B¡ rBtC|1|||$d9\}O}P}(})}*}2}3|2d d
df |2 Dd
¡ }Q}2|( D¡ |) D¡ |Q D¡ |2 D¡ f\}+},}-}.tEjF|1d GtEjH¡| d:}Rn
t )d
¡}Rd;}St |Sd< |R I¡ |+|,|-|.f ¡ |
s| d=k râ|sâ| d
krât/|1rât"|3D ]>\}T}Ut |S|$|U |R|U |(|T |)|T |Q|T |2|T f ¡ q¢tJ fd>d?|'D }V|s |d||f}It d@|I |V ¡ |rD|#jK|t'|$ L¡ dA | <dB¡ | rt/|0 r|d k r|tt|t'rt|d n|j:ndC}Wtt| dDdE¡dF }Xt||W dG }Yt dH|Y dI¡ tM|YdJ}NtN O|0|N¡ W 5 Q R X ztPdKg ddLlQmR}Z ddMlSmT}[ |Z|X}\|\ U|Y¡}D|[|\|DdN}]| r<dOd* |jVjWD |]jX_Y|] Z¡ |] [¡ |] \¡ |]j]d d+
\}.}-W n4 t^k
r }^ zt dP|^ ¡ W 5 d }^~^X Y nX | ¡ | sò| rÔdQt/t'| _dR¡ dS|d ndC}&t dTt dU| |& ¡ tE )| ¡|. }_t"|3D ]\}T}U|2|T |_|U<
q|+|,|-|.f|/ 8¡ t/| 0¡ |_|VfS )VNTFrS )Ú
batch_size)Úexist_okr[ )Úparentsri )rL Údnn)Úsr z2Forcing --batch-size 1 square inference shape(1,3,ú,z) for non-PyTorch backendsrd zcoco/val2017.txtÚncg à?gffffffî?é
r6 )ÚimgszÚhalfÚspeedg ©Útrainrd Útestz: )ÚpadÚrectÚprefixr )rn c S s i | ]\}}||qS r1 r1 )r9 ÚkÚvr1 r1 r2 Ú
<dictcomp> s zrun.<locals>.<dictcomp>Únamesiè z %20s%11s%11s%11s%11s%11s%11s)ZClassZImagesZLabelsÚPÚRzmAP@.5z
mAP@.5:.95)rL z {l_bar}{bar:10}{r_bar}{bar:-10b})ÚdescÚ
bar_format)Ú non_blockingéÿ )Úaugmentrd c S s g | ] }| ¡ qS r1 )Úfloatr8 r1 r1 r2 r; ¸ s zrun.<locals>.<listcomp>r4 c s, g | ]$} d d
df |kdd
f qS )Nr r r1 )r9 Úi)Útargetsr1 r2 r; ¼ s )r[ Z multi_labelZagnostic)rK r5 r ú.txt)r- Zon_val_image_endZ val_batchz _labels.jpg)ÚtargetÚargsÚdaemonz _pred.jpgc S s g | ]}t |d ¡qS )r )rV Ú concatenater8 r1 r1 r2 r; ï s )ÚplotÚsave_dirr| )Ú minlengthz$%20s%11i%11i%11.3g%11.3g%11.3g%11.3gÚallé2 c 3 s | ]}| d V qdS )g @@Nr1 r8 )Úseenr1 r2 Ú <genexpr> s zrun.<locals>.<genexpr>zKSpeed: %.1fms pre-process, %.1fms inference, %.1fms NMS per image at shape )r r| Z
on_val_endre rE z../cocoz"annotations/instances_val2017.jsonz_predictions.jsonz&
Evaluating pycocotools mAP... saving ú...ÚwZ pycocotools)ÚCOCO)ÚCOCOevalr= c S s g | ]}t t|jqS r1 )rA r r? r8 r1 r1 r2 r; s zpycocotools unable to run: r! z labels/*.txtz labels saved to zResults saved to Úbold)`ÚnextÚ
parametersrL Útyperq r r r r Úmkdirr ÚstrideÚptÚenginer Úmodelrh r" r Úinfor
ÚevalÚ
isinstanceÚgetÚstrÚendswithrA ÚlinspacerY ÚnumelÚwarmupr r r Ú enumerateÚhasattrr| Úmoduler ÚlistÚrangerN r r r, rX r r( r$ rC rO Úcloner r rQ r_ rS r3 r? rJ Úrunr r Ústartr rB Úanyr ÚmeanrV ÚbincountÚastypeÚint64ÚsumÚtupler Úvaluesr& ÚjsonÚdumpr Zpycocotools.cocor Zpycocotools.cocoevalr ZloadResÚdatasetZ img_filesÚparamsZimgIdsZevaluateÚ
accumulateÚ summarizeÚstatsÚ ExceptionÚglob)`ÚdataÚweightsrh rp Ú
conf_thresÚ iou_thresÚtaskrL Ú
single_clsr ÚverboseÚsave_txtÚ save_hybridr+ Ú save_jsonÚprojectÚnameri rq rk r Ú
dataloaderr ÚplotsÚ callbacksÚ compute_lossÚtrainingr r r Úis_cocorn r\ Zniourv Zconfusion_matrixr| rF rl ÚdtrH ÚrÚf1ÚmpÚmrZmap50ÚmapÚlossrD r¿ ÚapZap_classÚpbarZbatch_iZimÚpathsÚshapesÚt1ÚnbÚ_ÚheightÚwidthÚt2ÚoutZ train_outÚlbÚt3ÚsiÚpredr[ ÚnlZtclsrE r, r* ZtboxZlabelsnr] r0 ÚtpÚfpZap50ÚntZpfr
ÚcÚtr Z anno_jsonZ pred_jsonr r Úannor¡ ÚeÚmapsr1 )r r r2 r¯ S s"
ÿÿ
(*
0 (4$ 6 $ $
*8
,
",r¯ c C s° t ¡ } | jdttd dd | jddttd dd | jd
td d d | jd
ddtddd | jdtddd | jdtddd | jdddd | jd d d d | jd d d!d" | jd#d d$d" | jd%d d&d" | jd'd d(d" | jd)d d*d" | jd+d d,d" | jd-d d.d" | jd/td0 d1d | jd2d3d1d | jd4d d5d" | jd6d d7d" | jd8d d9d" | ¡ }t|j |_ | j
|j d:¡O _
| j |j
O _ ttj| |S );Nz--datazdata/coco128.yamlzdataset.yaml path)r ÚdefaultÚhelpz --weightsú+z
yolov5s.ptzmodel.pt path(s))Únargsr ró rô z --batch-sizer` z
batch sizez--imgszz--imgz
--img-sizera zinference size (pixels)z --conf-thresrb zconfidence thresholdz --iou-thresrc zNMS IoU thresholdz--taskrd z train, val, test, speed or study)ró rô z--devicere z%cuda device, i.e. 0 or 0,1,2,3 or cpuz --single-clsÚ
store_truez treat as single-class dataset)Úactionrô z --augmentzaugmented inferencez --verbosezreport mAP by classz
--save-txtzsave results to *.txtz
--save-hybridz-save label+prediction hybrid results to *.txtz --save-confz%save confidences in --save-txt labelsz --save-jsonz save a COCO-JSON results filez --projectrf zsave to project/namez--namerg z
--exist-okz*existing project/name ok, do not incrementz--halfz!use FP16 half-precision inferencez--dnnz!use OpenCV DNN for ONNX inferencez coco.yaml)ÚargparseÚArgumentParserÚ add_argumentr¤ ÚROOTrA r Ú
parse_argsr
r rË r¥ rÉ rÊ r ÚFILEr? )ÚparserÚoptr1 r1 r2 Ú parse_opt1 s6 r c C sp t td dd | jdkrJ| jdkr8t d| j d¡ tf t| n"t| j t
r\| j n| j g}d| _ | jd kr¨d
\| _| _ | _
|D ] | _ tf t| d d i qnÄ| jd
krl|D ]| _ dt| jj dt| j j d}t
tdddg }}|D ]J| _t d| d| j d¡ tf t| d d i\}}}| || ¡ qútj||dd q¸t d¡ t|d d S )Nzrequirements.txt)Z tensorboardÚthop)Ú requirementsÚexcluders rb z WARNING: confidence threshold z* >> 0.001 will produce invalid mAP values.Trr )g Ð?gÍÌÌÌÌÌÜ?FrÏ FZstudyZstudy_rá r é i é z
Running z --imgsz r z%10.4g)Úfmtz zip -r study.zip study_*.txt)r: )r rü rÆ rÄ r r r¯ Úvarsr¢ rà r¬ rq rÅ rË r r r? r rp rC rV ÚsavetxtÚosÚsystemr )r rà r0 r: ÚyrÕ rá rï r1 r1 r2 ÚmainO s,
"
r
Ú__main__)?Ú__doc__rù r¹ r
ÚsysÚpathlibr Ú threadingr rT rV r" r Ú__file__Úresolverþ rj rü r¤ rE rC ÚrelpathÚcwdZ
models.commonr Úutils.callbacksr Úutils.datasetsr Ú
utils.generalr r r
r r r
r r r r r r r r Ú
utils.metricsr r Ú utils.plotsr r r Úutils.torch_utilsr r r3 rJ r_ Úno_gradr¯ r r
Ú__name__r r1 r1 r1 r2 Ú<module> sr
@
ç ^