81 lines
10 KiB
Plaintext
81 lines
10 KiB
Plaintext
|
<03>
|
|||
|
|
<EFBFBD><[c @ s<> d d l Z d d l Z d d l Z d d l m Z m Z d d l m Z d <00> Z d <00> Z
|
|||
|
|
d d d d d d d
|
|||
|
|
<00> Z d <00> Z
d <00> Z d
<00> Z d <00> Z d <00> Z d <00> Z d <00> Z d <00> Z d d <00> Z d S( i<><69><EFBFBD><EFBFBD>Ni ( t BoundBoxt bbox_iou( t expitc C s
|
|||
|
|
t | <00> S( N( R ( t x( ( sH /home/dlsaavedra/Desktop/Tesis/8.-YOLO/keras-yolo3-master/utils/utils.pyt _sigmoid s c C sA y t j | <00> Wn) t k
|
|||
|
|
r< t j j | <00> s= <00> q= n Xd S( N( t ost makedirst OSErrort patht isdir( R ( ( sH /home/dlsaavedra/Desktop/Tesis/8.-YOLO/keras-yolo3-master/utils/utils.pyR
|
|||
|
|
s
|
|||
|
|
g <00>?g<><67><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>?i<> c # C s<> g t | j <00> <00> D]+ } g t | j <00> <00> D] } d ^ q, ^ q }
|
|||
|
|
g t | j <00> <00> D]+ } g t | j <00> <00> D] } d ^ qp ^ qW } x<>t | j <00> <00> D]<5D>} | j | <00> g } t | | | | | j <00> | | <00> d }
t j g |
D] } | j <00> ^ q<> <00> } t j g |
D] } | j
|
|||
|
|
^ q<01> } t |
<00> d k r<>t j g |
D]- } | j | j
| j | j | j <00> g ^ qG<01> }
n t j g g <00> }
t j | <0B> } | | } |
| }
x> t | j <00> <00> D]* } |
| | k d d <00> f |
|
|||
|
|
| | <q<>W| j | <00> } xT t | j <00> <00> D]@ } | | d d <00> d f | k d d <00> f j <00> | | | <qWq<57> Wi } x^t | j <00> <00> D]J} t j d <00> } t j d <00> } t j d <00> } d } xdt | j <00> <00> D]P} |
|
|||
|
|
| | } | | | } | | j d 7} g } x| D] } t j | | d <19> } | j d d k r_t j | d <00> } t j | d <00> } qn t t j | d d <00>| <00> } t j | d d <00>} | d | f } | | k r<>| | k r<>t j | d <00> } t j | d <00> } | j | <00> qt j | d <00> } t j | d <00> } qWq<57>W| d k r6d | | <qzn t j | <0B> } | | } | | } t j | <00> } t j | <00> } | | } | t j | | t j t j <00> j <00> }! t | |! <00> }" |" | | <qzW| S(
|
|||
|
|
sh Evaluate a given dataset using a given model.
|
|||
|
|
code originally from https://github.com/fizyr/keras-retinanet
|
|||
|
|
|
|||
|
|
# Arguments
|
|||
|
|
model : The model to evaluate.
|
|||
|
|
generator : The generator that represents the dataset to evaluate.
|
|||
|
|
iou_threshold : The threshold used to consider when a detection is positive or negative.
|
|||
|
|
obj_thresh : The threshold used to distinguish between object and non-object
|
|||
|
|
nms_thresh : The threshold used to determine whether two detections are duplicates
|
|||
|
|
net_h : The height of the input image to the model, higher value results in better accuracy
|
|||
|
|
net_w : The width of the input image to the model
|
|||
|
|
save_path : The path to save images with visualized detections to.
|
|||
|
|
# Returns
|
|||
|
|
A dict mapping class names to mAP scores.
|
|||
|
|
i Ni g i t axis( i ( i ( i ( t ranget sizet num_classest Nonet
|
|||
|
|
load_imaget get_yolo_boxest get_anchorst npt arrayt get_scoret labelt lent xmint ymint xmaxt ymaxt argsortt load_annotationt copyt zerost shapet appendt compute_overlapt expand_dimst argmaxt cumsumt maximumt finfot float64t epst
|
|||
|
|
compute_ap(# t modelt generatort
iou_thresholdt
|
|||
|
|
obj_thresht
|
|||
|
|
nms_thresht net_ht net_wt save_patht jt it all_detectionst all_annotationst raw_imaget
|
|||
|
|
pred_boxest boxt scoret pred_labelst
|
|||
|
|
score_sortR t annotationst average_precisionst false_positivest true_positivest scorest num_annotationst
|
|||
|
|
detectionst detected_annotationst dt overlapst assigned_annotationt max_overlapt indicest recallt precisiont average_precision( ( sH /home/dlsaavedra/Desktop/Tesis/8.-YOLO/keras-yolo3-master/utils/utils.pyt evaluate sr DD((%F
|
|||
|
|
|
|||
|
|
(B
|
|||
|
|
|
|||
|
|
|
|||
|
|
|
|||
|
|
)c C sJ t | <00> | t | <00> | k r7 | } | | | } n | } | | | } x<> t t | <00> <00> D]<5D> } | | d | t | <00> | } } | | d | t | <00> | }
|
|||
|
|
} t | | j | | | <14> | | _ t | | j | | | <14> | | _ t | | j |
|
|||
|
|
| | <14> | | _ t | | j |
|
|||
|
|
| | <14> | | _ q^ Wd S( Ng @( t floatR R t intR R R R ( t boxest image_ht image_wR/ R0 t new_wt new_hR3 t x_offsett x_scalet y_offsett y_scale( ( sH /home/dlsaavedra/Desktop/Tesis/8.-YOLO/keras-yolo3-master/utils/utils.pyt correct_yolo_boxes<65> s ##&&&c
|
|||
|
|
C s t | <00> d k r( t | d j <00> } n d Sx<53> t | <00> D]<5D> } t j g | D] } | j | ^ qL <00> } x<> t t | <00> <00> D]<5D> } | | } | | j | d k r<> q| n x\ t | d t | <00> <00> D]A } | | } t | | | | <19> | k r<> d | | j | <q<> q<> Wq| Wq9 Wd S( Ni i ( R t classesR R R R (
|
|||
|
|
RO R. t nb_classt cR8 t sorted_indicesR3 t index_iR2 t index_j( ( sH /home/dlsaavedra/Desktop/Tesis/8.-YOLO/keras-yolo3-master/utils/utils.pyt do_nms<6D> s *
|
|||
|
|
|
|||
|
|
c C s^ | j d \ } } d } | j | | | d f <00> } | j d d } g } t | d d d <00> f <19> | d d d <00> f <t | d <19> | d
|
|||
|
|
<| d d t j f t | d d d <00> f <19> | d d d <00> f <| d d d <00> f c | d d d <00> f | k 9<xYt | | <14> D]G}
|
|||
|
|
|
|
|||
|
|
| } |
|
|||
|
|
| } x*t | <00> D] }
| | | |
d f } | | k rdq6n | | | |
d d <00> f \ } } } } | | | } | | | } | d |
d t j | <00> | } | d |
d t j | <00> | } | | | |
d d <00> f } t | | d | | d | | d | | d | | <00> } | j | <00> q6WqW| S( Ni i i<><69><EFBFBD><EFBFBD>i .i i i ( .i ( .i ( .i (
|
|||
|
|
R t reshapeR R t newaxist _softmaxR t expR R ( t netoutt anchorsR- R/ R0 t grid_ht grid_wt nb_boxRZ RO R3 t rowt colt bt
|
|||
|
|
objectnessR t yt wt hRY R8 ( ( sH /home/dlsaavedra/Desktop/Tesis/8.-YOLO/keras-yolo3-master/utils/utils.pyt
decode_netout<75> s2 ,A2
|
|||
|
|
|
|||
|
|
(## ;c C s | j \ } } } t | <00> | t | <00> | k rI | | | } | } n | | | } | } t j | d d <00> d d <00> d d d <00> f d | | f <00> } t j | | d f <00> d } | | | | d | | d <1A> | | d | | d <1A> d d <00> f <t j | d <00> } | S( Ni<4E><69><EFBFBD><EFBFBD>g <00>o@i g <00>?i i ( R RM t cv2t resizeR t onesR" ( t imageR/ R0 RS RR t _t resizedt new_image( ( sH /home/dlsaavedra/Desktop/Tesis/8.-YOLO/keras-yolo3-master/utils/utils.pyt preprocess_input<75> s > Ec C s | d S( Ng <00>o@( ( Rt ( ( sH /home/dlsaavedra/Desktop/Tesis/8.-YOLO/keras-yolo3-master/utils/utils.pyt normalize<7A> s c C sY | d j \ } } } t | <00> }
|
|||
|
|
t j |
|
|||
|
|
| | d f <00> } x. t |
|
|||
|
|
<00> D] } t | | | | <00> | | <qJ W| j | <00> }
d g |
|
|||
|
|
} x<> t |
|
|||
|
|
<00> D]<5D> } |
d | |
d | |
d | g } g } xW t t | <00> <00> D]C } | d | d d | d !} | t | | | | | | <00> 7} q<> Wt | | | | | <00> t
|
|||
|
|
| | <00> | | | <q<> W| S( Ni i i i i ( R R R R R Rx t predict_on_batchR Rp RX R_ ( R* t imagesR/ R0 Re R- R. RP RQ Ru t nb_imagest batch_inputR3 t batch_outputt batch_boxest yolosRO R2 t yolo_anchors( ( sH /home/dlsaavedra/Desktop/Tesis/8.-YOLO/keras-yolo3-master/utils/utils.pyR <00> s"
' $
c C s
| d d <00> d f | d d <00> d f | d d <00> d f | d d <00> d f } t j t j | d d <00> d f d d <00>| d d <00> d f <19> t j t j | d d <00> d f d <00> | d d <00> d f <19> } t j t j | d d <00> d f d d <00>| d d <00> d f <19> t j t j | d d <00> d f d <00> | d d <00> d f <19> } t j | d <00> } t j | d <00> } t j | d d <00> d f | d d <00> d f | d d <00> d f | d d <00> d f d d <00>| | | } t j | t j t <00> j <00> } | | } | | S( s
|
|||
|
|
Code originally from https://github.com/rbgirshick/py-faster-rcnn.
|
|||
|
|
Parameters
|
|||
|
|
----------
|
|||
|
|
a: (N, 4) ndarray of float
|
|||
|
|
b: (K, 4) ndarray of float
|
|||
|
|
Returns
|
|||
|
|
-------
|
|||
|
|
overlaps: (N, K) ndarray of overlap between boxes and query_boxes
|
|||
|
|
Ni i i i R
|
|||
|
|
( R t minimumR" R% R& RM R( ( t aRk t areat iwt iht uat intersection( ( sH /home/dlsaavedra/Desktop/Tesis/8.-YOLO/keras-yolo3-master/utils/utils.pyR! s R}}m
|
|||
|
|
c C s<> t j d g | d g f <00> } t j d g | d g f <00> } xG t | j d d d <00> D], } t j | | d | | <19> | | d <qV Wt j | d | d k <00> d } t j | | d | | | | d <14> } | S( sI Compute the average precision, given the recall and precision curves.
|
|||
|
|
Code originally from https://github.com/rbgirshick/py-faster-rcnn.
|
|||
|
|
|
|||
|
|
# Arguments
|
|||
|
|
recall: The recall curve (list).
|
|||
|
|
precision: The precision curve (list).
|
|||
|
|
# Returns
|
|||
|
|
The average precision as computed in py-faster-rcnn.
|
|||
|
|
g g <00>?i i i<><69><EFBFBD><EFBFBD>( R t concatenateR R R% t wheret sum( RI RJ t mrect mpreR3 t ap( ( sH /home/dlsaavedra/Desktop/Tesis/8.-YOLO/keras-yolo3-master/utils/utils.pyR) $ s *!+c C sB | t j | | d t <00>} t j | <00> } | | j | d t <00>S( Nt keepdims( R t amaxt TrueRc R<> ( R R
|
|||
|
|
t e_x( ( sH /home/dlsaavedra/Desktop/Tesis/8.-YOLO/keras-yolo3-master/utils/utils.pyRb ? s ( Rq t numpyR R t bboxR R t
scipy.specialR R R R RL RX R_ Rp Rx Ry R R! R) Rb ( ( ( sH /home/dlsaavedra/Desktop/Tesis/8.-YOLO/keras-yolo3-master/utils/utils.pyt <module> s* l (
|