Can we predict not to use the annotations_test.csv file

[yc139464127]

Can you provide the code for the third step test, that is, you do not need to use the annotations_test.csv file, you can directly read the pictures under the images file to make predictions, then generate a csv file and save the resulting picture

[yc139464127]

That is to do inference without evaluation, and then no annotations_test.csv file

[eg4000]

Usage

The results are saved in CSV format in the "results" folder and drawn in "res_images_iou" folder.

[yc139464127]

Usage

The results are saved in CSV format in the "results" folder and drawn in "res_images_iou" folder.

I know this, but the prediction part of your code uses the annotations_train.csv file for evaluation, and at the end there will be two color boxes: green and red. Now what I want is to directly input a picture and then make predictions. The picture obtained is only the target frame and the csv file. Do you have such a code?

[eg4000]

Just comment out the parts that depend on the GT file.

You may find this useful as well.

[yc139464127]

thanks

[danbader88]

@yc139464127 Can you please share exactly what did you do?
I also want to make predictions on test images directly.

[blazpotocnik97]

Hi, I tried using this link you provided https://github.com/yfe404/SKU110K_code/blob/master/examples/retinanet.ipynb

But it is not working for me for some reason. I edited file paths and that was it. Script runs but I don't get any boxes on the output picture. Any idea what I'm doing wrong?

[Nhatcao99]

thanks

Hello, I haven't run the code yet but as mentioned in the paper, it also perform well in counting. Did the prediction part you manage to run consist of counting result??
THank you.

[webmonki]

Just comment out the parts that depend on the GT file.

You may find this useful as well.

@eg4000
What exactly do you mean by "GT file"?

[JavierClearImageAI]

my wrapper from the link you shared.

NOTE: I store it where predict.py is and run it by doing
'cd repo; $ python object_detector_retinanet/keras_retinanet/bin/our_predictor.py'

wrapper 1 (simple):

# import keras
import keras
# import keras_retinanet
from object_detector_retinanet.keras_retinanet import models
from object_detector_retinanet.keras_retinanet.utils.image import read_image_bgr, preprocess_image, resize_image
from object_detector_retinanet.keras_retinanet.utils.visualization import draw_box, draw_caption
from object_detector_retinanet.keras_retinanet.utils.colors import label_color
# import for EM Merger and viz
from object_detector_retinanet.keras_retinanet.utils import EmMerger
from object_detector_retinanet.utils import create_folder, root_dir
# import miscellaneous modules
import matplotlib.pyplot as plt
import cv2
import os
import numpy as np
import time
from tqdm import tqdm_notebook
# set tf backend to allow memory to grow, instead of claiming everything
import tensorflow as tf
​
def get_session():
    config = tf.ConfigProto()
    config.gpu_options.allow_growth = True
    return tf.Session(config=config)
​​
# use this environment flag to change which GPU to use
#os.environ["CUDA_VISIBLE_DEVICES"] = "1"
​
# set the modified tf session as backend in keras
keras.backend.tensorflow_backend.set_session(get_session())
​
# ### Load RetinaNet model
# # adjust this to point to your downloaded/trained model
# # models can be downloaded here: https://github.com/fizyr/keras-retinanet/releases
# model_path = os.path.join('..', 'retinanet_inference.h5')
# 
# # load retinanet model
# model = models.load_model(model_path, backbone_name='resnet50')
# 
# # load label to names mapping for visualization purposes
# labels_to_names = {0: 'object'}
​
### Load Retinanet + IoU Model
model_path = '/home/ubuntu/Documents/iou_resnet50_csv_06.h5'
model = models.load_model(model_path, backbone_name='resnet50', convert=1, nms=False)
# model = models.load_model(model_path, backbone_name='resnet50')
# load label to names mapping for visualization purposes
labels_to_names = {0: 'object'}
​
# load image
image_path = '/home/ubuntu/Documents/SKU110K/beers/PIWA_AR3006_F00007_468943_PIWO 1.jpg'
image = read_image_bgr(image_path)
​
# for filtering predictions based on score (objectness/confidence)
threshold = 0.3
​
# copy to draw on
draw = image.copy()
draw = cv2.cvtColor(draw, cv2.COLOR_BGR2RGB)
​
# preprocess image for network
image = preprocess_image(image)
image, scale = resize_image(image)
​
# Run inference
boxes, hard_scores, labels, soft_scores = model.predict_on_batch(np.expand_dims(image, axis=0))
​
#########################
hard_score_rate=.3
max_detections = 9999
​
soft_scores = np.squeeze(soft_scores, axis=-1)
soft_scores = hard_score_rate * hard_scores + (1 - hard_score_rate) * soft_scores
​
# correct boxes for image scale
boxes /= scale
​
# select indices which have a score above the threshold
indices = np.where(hard_scores[0, :] > threshold)[0]
​
# select those scores
scores = soft_scores[0][indices]
hard_scores = hard_scores[0][indices]
​
# find the order with which to sort the scores
scores_sort = np.argsort(-scores)[:max_detections]
​
# select detections
image_boxes = boxes[0, indices[scores_sort], :]
image_scores = scores[scores_sort]
image_hard_scores = hard_scores[scores_sort]
image_labels = labels[0, indices[scores_sort]]
image_detections = np.concatenate(
    [image_boxes, np.expand_dims(image_scores, axis=1), np.expand_dims(image_labels, axis=1)], axis=1)
results = np.concatenate(
    [image_boxes, np.expand_dims(image_scores, axis=1), np.expand_dims(image_hard_scores, axis=1),
     np.expand_dims(image_labels, axis=1)], axis=1)
filtered_data = EmMerger.merge_detections(image_path, results)
filtered_boxes = []
filtered_scores = []
filtered_labels = []
​
csv_data_lst = []
csv_data_lst.append(['image_id', 'x1', 'y1', 'x2', 'y2', 'confidence', 'hard_score'])
​
for ind, detection in filtered_data.iterrows():
    box = np.asarray([detection['x1'], detection['y1'], detection['x2'], detection['y2']])
    filtered_boxes.append(box)
    filtered_scores.append(detection['confidence'])
    filtered_labels.append('{0:.2f}'.format(detection['hard_score']))
    row = [image_path, detection['x1'], detection['y1'], detection['x2'], detection['y2'],
           detection['confidence'], detection['hard_score']]
    csv_data_lst.append(row)
​
for box, score, label in zip(filtered_boxes, filtered_scores, filtered_labels):
    # scores are sorted so we can break
    if score < threshold:
        break
        
    color = [31, 0, 255]#label_color(label) ## BUG HERE LABELS ARE FLOATS SO COLOR IS HARDCODED 
    
    b = box.astype(int)
    draw_box(draw, b, color=color)
    
    caption = "{} {:.3f}".format(labels_to_names[0], score)
    draw_caption(draw, b, caption)
​
plt.figure(figsize=(20, 20))
plt.axis('off')
plt.imshow(draw)
plt.savefig('/home/ubuntu/Documents/SKU110K/beers/PIWA_AR3006_F00007_468943_PIWO 1_predicted.jpg', dpi=100)
# plt.show()

---

---

Wrapper 2 (more functional):

# import keras
import keras

# import keras_retinanet
from object_detector_retinanet.keras_retinanet import models
from object_detector_retinanet.keras_retinanet.utils.image import read_image_bgr, preprocess_image, resize_image
from object_detector_retinanet.keras_retinanet.utils.visualization import draw_box, draw_caption
from object_detector_retinanet.keras_retinanet.utils.colors import label_color

# import for EM Merger and viz
from object_detector_retinanet.keras_retinanet.utils import EmMerger
from object_detector_retinanet.utils import create_folder, root_dir
# import miscellaneous modules
import matplotlib.pyplot as plt
import cv2
import os
from os.path import join
import shutil as sh
import numpy as np
import time
from tqdm import tqdm
from PIL import Image

# set tf backend to allow memory to grow, instead of claiming everything
import tensorflow as tf

def get_session():
    config = tf.ConfigProto()
    config.gpu_options.allow_growth = True
    return tf.Session(config=config)

# use this environment flag to change which GPU to use
#os.environ["CUDA_VISIBLE_DEVICES"] = "1"

# set the modified tf session as backend in keras
keras.backend.tensorflow_backend.set_session(get_session())


class load_our_config:
    def __init__(self):
        ### img_path or folder
        self.img_dir = '/home/ubuntu/Documents/SKU110K/beers/'
        self.predictions_dir = join(self.img_dir, 'predictions')        
        ### delete old predictions if the dir exists
        try:
            sh.rmtree(self.predictions_dir)
        except:
            pass

        self.image_name = False # 'PIWA_AR3006_F00007_468943_PIWO 1.jpg' # set to False to use whole directory
        if self.image_name:
            self.images_paths = [join(self.img_dir, self.image_name)]  # add it as list so we can iterate despite format
        else:
            ### it's a directory. get files there (be sure to put only images!)
            self.images_paths = [join(self.img_dir,x) for x in os.listdir(self.img_dir)]
        print("self.images_paths: ", self.images_paths)
        
        ### create a new folder for predicted images
        os.mkdir(self.predictions_dir)
        self.model_path = '/home/ubuntu/Documents/iou_resnet50_csv_06.h5'
        self.backbone_name = 'resnet50'
        self.hard_score_rate=.3
        self.max_detections = 9999
        # for filtering predictions based on score (objectness/confidence)
        self.threshold = 0.3
        # load label to names mapping for visualization purposes
        self.labels_to_names = {0: 'object'}

def main(our_config):
    ### Load Retinanet + IoU Model
    our_config.model = models.load_model(our_config.model_path, backbone_name=our_config.backbone_name,
        convert=1, nms=False)
    # model = models.load_model(model_path, backbone_name='resnet50')

    for our_config.image_path in tqdm(our_config.images_paths):
        # load image
        our_config.image = read_image_bgr(our_config.image_path)
        predict(our_config)

def predict(our_config):
    # copy to draw on
    draw = our_config.image.copy()
    draw = cv2.cvtColor(draw, cv2.COLOR_BGR2RGB)
    
    # preprocess image for network
    image = preprocess_image(our_config.image)
    image, scale = resize_image(image)
    
    # Run inference
    boxes, hard_scores, labels, soft_scores = our_config.model.predict_on_batch(np.expand_dims(image, axis=0))
    #########################
    soft_scores = np.squeeze(soft_scores, axis=-1)
    soft_scores = our_config.hard_score_rate * hard_scores + (1 - our_config.hard_score_rate) * soft_scores
    # correct boxes for image scale
    boxes /= scale
    # select indices which have a score above the threshold
    indices = np.where(hard_scores[0, :] > our_config.threshold)[0]
    
    # select those scores
    scores = soft_scores[0][indices]
    hard_scores = hard_scores[0][indices]
    
    # find the order with which to sort the scores
    scores_sort = np.argsort(-scores)[:our_config.max_detections]
    
    # select detections
    image_boxes = boxes[0, indices[scores_sort], :]
    image_scores = scores[scores_sort]
    image_hard_scores = hard_scores[scores_sort]
    image_labels = labels[0, indices[scores_sort]]
    image_detections = np.concatenate(
        [image_boxes, np.expand_dims(image_scores, axis=1), np.expand_dims(image_labels, axis=1)], axis=1)
    results = np.concatenate(
        [image_boxes, np.expand_dims(image_scores, axis=1), np.expand_dims(image_hard_scores, axis=1),
         np.expand_dims(image_labels, axis=1)], axis=1)
    filtered_data = EmMerger.merge_detections(our_config.image_path, results)
    filtered_boxes = []
    filtered_scores = []
    filtered_labels = []
    
    csv_data_lst = []
    csv_data_lst.append(['image_id', 'x1', 'y1', 'x2', 'y2', 'confidence', 'hard_score'])
    
    for ind, detection in filtered_data.iterrows():
        box = np.asarray([detection['x1'], detection['y1'], detection['x2'], detection['y2']])
        filtered_boxes.append(box)
        filtered_scores.append(detection['confidence'])
        filtered_labels.append('{0:.2f}'.format(detection['hard_score']))
        row = [our_config.image_path, detection['x1'], detection['y1'], detection['x2'], detection['y2'],
               detection['confidence'], detection['hard_score']]
        csv_data_lst.append(row)
     
    
    for box, score, label in zip(filtered_boxes, filtered_scores, filtered_labels):
        # scores are sorted so we can break
        if score < our_config.threshold:
            break
            
        color = [31, 0, 255] #label_color(label) ## BUG HERE LABELS ARE FLOATS SO COLOR IS HARDCODED 
        
        b = box.astype(int)
        draw_box(draw, b, color=color)
        
        caption = "{} {:.3f}".format(our_config.labels_to_names[0], score)
        draw_caption(draw, b, caption)
    
    plt.figure(figsize=(20, 20))
    plt.axis('off')
    plt.imshow(draw)
    base_name = os.path.split(our_config.image_path)[1]
    predicted_img_path = join(our_config.predictions_dir, base_name)
    plt.savefig(predicted_img_path, bbox_inches='tight', pad_inches=0, dpi=220)


if __name__ == '__main__':
    our_config = load_our_config()
    # print(config.model_path)
    main(our_config)

[JavierClearImageAI]

PR added and approved by some developers, but not officially accepted yet.
https://github.com/eg4000/SKU110K_CVPR19/blob/56747143d8b6afcdf8864deffb6b8be45775caff/predict_only.py