Kot-Nak8
commented
4 years ago I may be interpreting your problem incorrectly, but I will report my results.
The code I've rewritten looks like this.
def init(self, **kwargs): self.dict.update(self._defaults) # set up default values self.dict.update(kwargs) # and update with user overrides self.class_names = self._get_class() self.anchors = self._get_anchors() self.sess = K.get_session() self.boxes, self.scores, self.classes ,self.out_model= self.generate()
def generate(self): .... .... .... return boxes, scores, classes, self.yolo_model.output
def detect_image(self, image): .... .... .... out_boxes, out_scores, out_classes, out_model = self.sess.run( [self.boxes, self.scores, self.classes, self.out_model], feed_dict={ self.yolo_model.input: image_data, self.input_image_shape: [image.size[1], image.size[0]], K.learning_phase(): 0 }) .... .... .... return image, out_model
The code I ran is as follows
import numpy as np import keras.backend as K from keras.models import load_model from keras.layers import Input from yolo import YOLO from PIL import Image import tensorflow as tf
image_path = '/Users/Users/Documents/keras-yolo3-master/test.png' image = Image.open(image_path) image_output , Model_output = YOLO().detect_image(image) print(Model_output)
The results were as follows
Found 5 boxes for img dog 1.00 (16, 204) (177, 487) truck 0.97 (327, 78) (504, 152) car 0.11 (325, 86) (510, 155) bicycle 0.33 (5, 122) (137, 338) bicycle 0.99 (0, 109) (426, 399) [array([[[[-7.09539771e-01, 4.01784360e-01, 1.45798981e-01, ..., -5.06193352e+00, -5.41232395e+00, -5.28670883e+00], [ 1.30866611e+00, 1.45839393e+00, -2.95680016e-02, ..., -5.93569422e+00, -7.04240608e+00, -7.55034590e+00], [ 3.97958338e-01, 1.28708482e+00, 2.58324027e-01, ..., -7.73941612e+00, -8.17567921e+00, -8.91163921e+00], ..., [ 1.40693784e-01, -5.44680655e-01, 4.12246287e-01, ..., -8.56248188e+00, -8.92824364e+00, -1.04769955e+01], [-8.13400507e-01, -1.18537128e+00, 2.49891207e-02, ..., -4.21445274e+00, -6.39804935e+00, -7.25709629e+00], [ 3.22407544e-01, -5.17006218e-01, 4.55918819e-01, ..., -3.19745970e+00, -4.74484110e+00, -4.39262247e+00]]]], dtype=float32), .... .... .... .... ....
I would like to get a feature map as well, but I don't know how to go about it from here. I don't know if my response is correct, but I hope it helps.
Hi, I am now doing the thesis about the YOLO. I want to get all the features of the three output layers of YOLO when giving the images into it, and the size of y1,y2,y3 should be (sample_num, 13, 13, num_anchors(5 + num_classes)=255(default)),(sample_num, 26, 26, num_anchors(5 + num_classes)=255(default),(sample_num, 52, 52, num_anchors*(5 + num_classes)=255(default) separately. However after testing for one image, I got following output, which didn't show the exact shape of the output layers.
`model_data/yolo.h5 model, anchors, and classes loaded. (416, 416, 3) Found 3 boxes for img ('dog 0.99', (128, 224), (314, 537)) ('truck 0.91', (475, 85), (689, 170)) ('bicycle 0.99', (162, 119), (565, 441)) 4.62596893311
Tensor("input_1:0", shape=(?, ?, ?, 3), dtype=float32) [<tf.Tensor 'conv2d_59/BiasAdd:0' shape=(?, ?, ?, 255) dtype=float32>, <tf.Tensor 'conv2d_67/BiasAdd:0' shape=(?, ?, ?, 255) dtype=float32>, <tf.Tensor 'conv2d_75/BiasAdd:0' shape=(?, ?, ?, 255) dtype=float32>]`
I got stuck in this problem for several days. Can anybody give some advice on how to solve this problem? Many thanks in advance!!!
My code is following:
`import numpy as np import keras.backend as K from keras.models import load_model from keras.layers import Input from yolo import YOLO from PIL import Image import tensorflow as tf
weights_path='/home/jichen/keras-yolo3/model_data/yolo.h5' image_path = '/home/jichen/Desktop/dog.jpg' image = Image.open(image_path) yolo_guo = YOLO() Model = yolo_guo.detect_image(image, is_tsne=True) sess=tf.Session() print Model.input print Model.output`
and for function detect_imiage, I want to return the model of yolo, so just modify a little bit:
if is_tsne: return self.yolo_model else: return image