Deep Learning for Tomato Disease Detection with YOLOv8

Engineering Technology & Applied Science Research, Journal Year: 2024, Volume and Issue: 14(2), P. 13584 - 13591

Published: April 2, 2024

Tomato production plays a crucial role in Saudi Arabia, with significant yield variations due to factors such as diseases. While automation offers promising solutions, accurate disease detection remains challenge. This study proposes deep learning approach based on the YOLOv8 algorithm for automated tomato detection. Augmenting an existing Roboflow dataset, model achieved overall accuracy of 66.67%. However, class-specific performance varies, highlighting challenges differentiating certain Further research is suggested, focusing data balancing, exploring alternative architectures, and adopting disease-specific metrics. work lays foundation robust system improve crop yields, quality, sustainable agriculture Arabia.

Language: Английский

---

Improved Tomato Disease Detection with YOLOv5 and YOLOv8

Engineering Technology & Applied Science Research, Journal Year: 2024, Volume and Issue: 14(3), P. 13922 - 13928

Published: June 1, 2024

This study delves into the application of deep learning for precise tomato disease detection, focusing on four crucial categories: healthy, blossom end rot, splitting rotation, and sun-scaled rotation. The performance two lightweight object detection models, namely YOLOv5l YOLOv8l, was compared a custom dataset. Initially, both models were trained without data augmentation to establish baseline. Subsequently, diverse techniques obtained from Roboflow significantly expand enrich dataset content. These aimed enhance models' robustness variations in lighting, pose, background conditions. Following augmentation, YOLOv8l re-trained their across all categories meticulously analyzed. After significant improvement accuracy observed highlighting its effectiveness bolstering ability accurately detect diseases. consistently achieved slightly higher YOLOv5l, particularly when excluding images evaluation.

Language: Английский

---

Comparison of YOLOv5 and YOLOv6 Models for Plant Leaf Disease Detection

Engineering Technology & Applied Science Research, Journal Year: 2024, Volume and Issue: 14(2), P. 13714 - 13719

Published: April 2, 2024

Deep learning is a concept of artificial neural networks and subset machine learning. It deals with algorithms that train process datasets to make inferences for future samples, imitating the human from experiences. In this study, YOLOv5 YOLOv6 object detection models were compared on plant dataset in terms accuracy time metrics. Each model was trained obtain specific results mean Average Precision (mAP) training time. There no considerable difference mAP between both models, as their close. YOLOv5, having 63.5% mAP, slightly outperformed YOLOv6, while 49.6% mAP50-95, better than YOLOv5. Furthermore, data shorter since it has fewer parameters.

Language: Английский

---

Q_YOLOv5m: A Quantization-based Approach for Accelerating Object Detection on Embedded Platforms

Engineering Technology & Applied Science Research, Journal Year: 2025, Volume and Issue: 15(1), P. 19749 - 19755

Published: Feb. 2, 2025

The deployment of deep learning models on resource-constrained embedded platforms presents significant challenges due to limited computational power, memory, and energy efficiency. To address this issue, study proposes a novel quantization method tailored accelerate object detection using quantized version the YOLOv5m model, called Q_YOLOv5m. This reduces model's complexity memory footprint, allowing for faster inference lower power consumption, making it ideal real-time applications systems. approach incorporates advanced weight activation techniques balance performance with accuracy, dynamically adjusting precision based hardware capabilities. efficacy Q_YOLOv5m was confirmed, exhibiting substantial enhancements in speed reduction model size negligible loss accuracy. findings underscore capability edge applications, including autonomous vehicles, intelligent surveillance, IoT-based monitoring

Language: Английский

---

Low brightness PCB image enhancement algorithm for FPGA

Journal of Real-Time Image Processing, Journal Year: 2025, Volume and Issue: 22(2)

Published: March 10, 2025

Language: Английский

---

Efficient Hardware Accelerator and Implementation of JPEG 2000 MQ Decoder Architecture

Engineering Technology & Applied Science Research, Journal Year: 2024, Volume and Issue: 14(2), P. 13463 - 13469

Published: April 2, 2024

Due to the extensive use of multimedia technologies, there is a pressing need for advancements and enhanced efficiency in picture compression. JPEG 2000 standard aims meet needs encoding still pictures. an internationally recognized compressing images. It provides wide range features offers superior compression ratios interesting possibilities when compared traditional approaches. Nevertheless, MQ decoder presents substantial obstacle real-time applications. In order fulfill demands processing, it imperative meticulously devise high-speed architecture. This work novel architecture that both area-efficient, making comparable previous designs well-suited chip implementation. The design implemented using VHDL hardware description language synthesized with Xilinx ISE 14.7 Vivado 2015.1. implementation findings show functions at frequency 438.5 MHz on Virtex-6 757.5 Zync7000. For these particular frequencies, calculated frame rate 63.1 frames per second.

Language: Английский

---

A Children's Psychological and Mental Health Detection Model by Drawing Analysis based on Computer Vision and Deep Learning

Engineering Technology & Applied Science Research, Journal Year: 2024, Volume and Issue: 14(4), P. 15533 - 15540

Published: Aug. 2, 2024

Nowadays, children face different changes and challenges from an early age, which can have long-lasting impacts on them. Many struggle to express or explain their feelings thoughts properly. Due that fact, psychological mental health specialists found a way detect issues by observing analyzing signs in children’s drawings. Yet, this process remains complex time-consuming. This study proposes solution employing artificial intelligence analyze drawings provide diagnosis rates with high accuracy. While prior research has focused detecting through questionnaires, only one explored emotions children's positive negative feelings. A notable gap is the limited of specific issues, along promising accuracy detection results. In study, versions YOLO were trained dataset 500 drawings, split into 80% for training, 10% validation, testing. Each drawing was annotated more emotional labels: happy, sad, anxiety, anger, aggression. YOLOv8-cls, YOLOv9, ResNet50 used object classification, achieving accuracies 94%, 95.1%, 70.3%, respectively. YOLOv9 results obtained at epoch numbers large model sizes 5.26 MB 94.3 MB. YOLOv8-cls achieved most satisfying result, reaching 94% after 10 epochs compact size 2.83 MB, effectively meeting study's goals.

Language: Английский

---

Model-based Design of a High-Throughput Canny Edge Detection Accelerator on Zynq-7000 FPGA

Engineering Technology & Applied Science Research, Journal Year: 2024, Volume and Issue: 14(2), P. 13547 - 13553

Published: April 2, 2024

This paper presents a novel approach for fast FPGA prototyping of the Canny edge detection algorithm using High-Level Synthesis (HLS) based on HDL Coder. Traditional RTL-based design methodologies implementing image processing algorithms FPGAs can be time-consuming and error-prone. HLS offers higher level abstraction, enabling designers to focus algorithmic functionality while tool automatically generates efficient hardware descriptions. advantage was exploited by in MATLAB/Simulink utilizing Coder convert it into synthesizable VHDL code. flow significantly reduces development time complexity compared traditional RTL approach. The experimental results showed that HLS-based detector achieved real-time performance Xilinx platform, showcasing effectiveness proposed applications.

Language: Английский

---