LiweiPE-Object_detection_drinks

This reasearch was develop to detect many soft drinks or snacks taking out from vending machine.

Configuration environment

conda activate tf1.15

pip install imageio

Dataset

Images annotations have to be changed to xmal format Whole raw dataset can be found at /data/vending Training dataset to detect one object can be found at /data/home/liwei/ImageAI/vending_machine Training dataset to detect more than two objects at the same time can be found at /data/home/liwei/ImageAI/vending_machine_dual

Training

The following example is used to detect different custom objects

from imageai.Detection.Custom import DetectionModelTrainer

trainer = DetectionModelTrainer()
trainer.setModelTypeAsYOLOv3()
trainer.setGpuUsage(train_gpus="0,1,2,3,4,5,6,7")
trainer.setDataDirectory(data_directory="vending_machine")
# trainer.setTrainConfig(object_names_array=["orange_juice"], batch_size=8, num_experiments=10,train_from_pretrained_model="pretrained-yolov3.h5")
trainer.setTrainConfig(object_names_array=["pepsi","water","orange_juice","cucumber_soda","C100_juice","pepsi_330","HongNiu","Wangzi_milk","Wanglaoji","Beibingyang","Asamu_milktea","Harbin_beer","Kangshifu_juice","Maidong_lime","Dongfang_greentea"],
                       batch_size=32, num_experiments=10,train_from_pretrained_model="vending_machine/models_all_25-11/detection_model-ex-010--loss-0007.615.h5")
trainer.trainModel()

Training file for one custom object detection

$ python vending_machine.py

Training file for multiple custom object detection

$ python vending_machine_dual.py

Test

$ python test_video.py

Example for testing video:

from imageai.Detection.Custom import CustomVideoObjectDetection
import os

execution_path = os.getcwd()
video_detector = CustomVideoObjectDetection()
video_detector.setModelTypeAsYOLOv3()
video_detector.setModelPath("vending_machine_dual/models1/detection_model-ex-039--loss-0011.630.h5")
video_detector.setJsonPath("vending_machine_dual/json1/detection_config.json")
video_detector.loadModel()

video_detector.detectObjectsFromVideo(input_file_path="01茅聬隆忙茠掳卯鈥斉犆┞惵ヂ徛ッ溍┞嶁劉卯藛鈩⒚?00氓搂拢卯鈥毬⒚ヂ磁?忙聬麓氓鲁掳莽卢鈧┞嵟捗モ€樎モ€⒙好宦?颅忙麓漏氓搂艩?00氓搂拢卯鈥毬⒚ヂ磁捗┞嶁劉氓鈥︹€γ€?avi",
                                          output_file_path=os.path.join(execution_path, "pepsi_kangshifu%"),
                                          frames_per_second=30,
                                          minimum_percentage_probability=50,
                                          log_progress=True)

Images samples detection

Single object: Orange juice 500ml

Multiple object: Pepsi 500ml, Cucumber soda 500ml, Water 500ml

Video detection: Pepsi 500ml

Citation

@misc {ImageAI, author = "Moses and John Olafenwa", title = "ImageAI, an open source python library built to empower developers to build applications and systems with self-contained Computer Vision capabilities", url = "https://github.com/OlafenwaMoses/ImageAI", month = "mar", year = "2018--" }