Automated Image Annotation based on YOLOv3

Tumas, Paulius; Serackis, Artūras

doi:10.1109/aieee.2018.8592167

Cited by 16 publications

(10 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Another example of a convolutional neural network-based thermal image annotation is Reference [ 28 ], but the authors focus only on a single class and do not provide any number-based evaluation of their approach.…”

Section: Discussionmentioning

confidence: 99%

See 1 more Smart Citation

Fully Automated DCNN-Based Thermal Images Annotation Using Neural Network Pretrained on RGB Data

Ligocki

Jelinek

Žalud

et al. 2021

Sensors

View full text Add to dashboard Cite

One of the biggest challenges of training deep neural network is the need for massive data annotation. To train the neural network for object detection, millions of annotated training images are required. However, currently, there are no large-scale thermal image datasets that could be used to train the state of the art neural networks, while voluminous RGB image datasets are available. This paper presents a method that allows to create hundreds of thousands of annotated thermal images using the RGB pre-trained object detector. A dataset created in this way can be used to train object detectors with improved performance. The main gain of this work is the novel method for fully automatic thermal image labeling. The proposed system uses the RGB camera, thermal camera, 3D LiDAR, and the pre-trained neural network that detects objects in the RGB domain. Using this setup, it is possible to run the fully automated process that annotates the thermal images and creates the automatically annotated thermal training dataset. As the result, we created a dataset containing hundreds of thousands of annotated objects. This approach allows to train deep learning models with similar performance as the common human-annotation-based methods do. This paper also proposes several improvements to fine-tune the results with minimal human intervention. Finally, the evaluation of the proposed solution shows that the method gives significantly better results than training the neural network with standard small-scale hand-annotated thermal image datasets.

show abstract

Section: Discussionmentioning

confidence: 99%

“…There are also propositions to use the networks pre-trained on RGB images to directly perform object detection on the thermal images [ 28 ], but this approach does not provide any benefit over the knowledge comprised in the neural network’s weights.…”

Section: Introductionmentioning

confidence: 99%

Fully Automated DCNN-Based Thermal Images Annotation Using Neural Network Pretrained on RGB Data

Ligocki

Jelinek

Žalud

et al. 2021

Sensors

View full text Add to dashboard Cite

show abstract

“…In order to better select the previous network, YOLOv3 inherits the YOLOv2 calculation anchor frame selection method and uses the K ‐means clustering method to train the bounding box. This method uses the IoU score as the final evaluation criterion and selects nine anchor points based on the average IoU to predict the bounding box thus achieving an improvement in precision [20]. The distance function formula used for clustering is

d false(box, centroid false) = 1 - IoU ()(, box, centroid)

…”

Section: Methodsmentioning

confidence: 99%

Research on automatic location and recognition of insulators in substation based on YOLOv3

Liu

Pei

et al. 2020

High Voltage

View full text Add to dashboard Cite

With the development of a smart grid, the automatic location of power equipment is becoming a trend. In this study, a method for automatic location identification and diagnosis of external power insulation equipment based on YOLOv3 is proposed. This deep learning algorithm is used to extract the characteristics of image data under the visible light channel of the insulator. It learns and trains the collected data to realise the rapid location identification and frame selection of the external insulation equipment and extract discharge characteristics of the target box under the ultraviolet channel. According to the number of photons and the spot area information, the operating status of the equipment is determined. The results show that the YOLOv3 algorithm with a training rate of 0.005 achieved a fast convergence of the location recognition model. The average recognition accuracy was 88.7% and the average detection time was 0.0182 s. The combination of visible light path insulator target recognition and ultraviolet light path diagnosis can realise a lean and intelligent diagnosis of power equipment. This method had good real-time performance, accuracy, and robustness to the background. It provides a new concept for intelligent diagnosis and location analysis of power equipment.

show abstract

“…Then, CAN data from UDP capture is copied by a thread locking semaphore. To minimize annotation work, we used the pre-annotation approach [36] based on the TINYv3 [37] neural network. The pre-annotation detector was trained by 50,000 of images scaled down to 640 × 480 resolution taken from SCUT dataset with ''People-?''…”

Section: B Data Recordingmentioning

confidence: 99%

Pedestrian Detection in Severe Weather Conditions

2020

View full text Add to dashboard Cite

Pedestrian detection has never been an easy task for computer vision and the automotive industry. Systems like the advanced driver-assistance system (ADAS) highly rely on far-infrared (FIR) data captured to detect pedestrians at nighttime. The recent development of deep learning-based detectors has proven the excellent results of pedestrian detection in perfect weather conditions. However, it is still unknown what the performance in adverse weather conditions is. In this paper, we introduce a 16-bit thermal data dataset called ZUT (Zachodniopomorski Uniwersytet Technologiczny) as having the widest variety of fine-grained annotated images captured in the four biggest European Union countries captured during severe weather conditions. We also provide a synchronized Controller Area Network (CAN bus) data, including driving speed, brake pedal status, and outside temperature for future ADAS system development. Furthermore, we have tested and provided 16-bit depth modifications for the YOLOv3 deep neural network (DNN) based detector, reaching a mean Average Precision (mAP) up to 89.1%. The ZUT dataset is published and publicly available at IEEE Dataport and Github.

show abstract

Automated Image Annotation based on YOLOv3

Cited by 16 publications

References 11 publications

Fully Automated DCNN-Based Thermal Images Annotation Using Neural Network Pretrained on RGB Data

Fully Automated DCNN-Based Thermal Images Annotation Using Neural Network Pretrained on RGB Data

Research on automatic location and recognition of insulators in substation based on YOLOv3

Pedestrian Detection in Severe Weather Conditions

Contact Info

Product

Resources

About