Inventory counting by a deep learning based object detection model in stock yard of an urban furniture manufacturer
Kent mobilyaları üretim firmasına ait stok sahasında derin öğrenme tabanlı bir nesne algılama modeli ile stok sayımı
- Tez No: 895626
- Danışmanlar: DR. ÖĞR. ÜYESİ ABDULLAH HULUSİ KÖKÇAM
- Tez Türü: Yüksek Lisans
- Konular: Endüstri ve Endüstri Mühendisliği, Industrial and Industrial Engineering
- Anahtar Kelimeler: Derin Öğrenme, Evrişimli Sinir Ağları, Nesne Algılama, YOLO, Envanter Sayımı, Envanter Sayım Araçları, Kent Mobilyaları Envanter sayımı
- Yıl: 2024
- Dil: İngilizce
- Üniversite: Sakarya Üniversitesi
- Enstitü: Fen Bilimleri Enstitüsü
- Ana Bilim Dalı: Endüstri Mühendisliği Ana Bilim Dalı
- Bilim Dalı: Endüstri Mühendisliği Bilim Dalı
- Sayfa Sayısı: Belirtilmemiş.
Özet
Özet yok.
Özet (Çeviri)
Inventory counting is a critical financial process for organizations, as it ensures accurate reporting of raw materials, components, and finished products within production and logistics operations. Precise inventory counts are essential for preventing financial losses that arise from discrepancies between recorded and actual quantities. Accurate counts help avoid issues such as overstocking, which leads to unnecessary expenses, and understocking, which can disrupt production processes. The use of computer vision in inventory counting offers a promising approach to enhancing accuracy and efficiency, addressing the limitations inherent in traditional methods. Manual inventory counting, while effective for small quantities of neatly arranged items, becomes increasingly problematic with larger and more complex stockpiles. The process is labor-intensive and time-consuming, particularly when dealing with extensive volumes of diverse products. Traditional methods such as Radio Frequency Identification (RFID) and barcode/QR code scanning face significant limitations. RFID signals can be obstructed by materials like concrete, and barcode/QR code labels are susceptible to weather conditions and adherence issues, leading to inaccuracies in counting and inefficiencies in inventory management. To address these challenges, this study explores the application of computer vision and deep learning techniques, specifically Convolutional Neural Networks (CNNs), to improve inventory counting. The methodology involved creating a unique dataset of images of selected products, categorized into four classes, and performing object labeling on hundreds of images. Data augmentation techniques were employed to enhance the robustness of the model. The YOLOv10 (You Only Look Once version 10) model was then trained using the augmented dataset to perform object detection and counting. The results of the study indicated Mean Average Precision(mAP) score of %86.9 that the YOLOv10 model successfully counted products in various scenarios, demonstrating the potential of computer vision to enhance inventory accuracy. Metrics and visualizations revealed that the model was effective in certain contexts, though limitations were observed, particularly in handling complex inventory scenes and specific product types. The study concludes that while computer vision and deep learning offer significant improvements over traditional methods, further refinements and additional research are needed to address the identified challenges and optimize inventory counting processes.
Benzer Tezler
- Karbonhidrat sayımı yapan Tip 1 diyabetik yetişkin bireylerin sağlıklı yeme takıntısı ve yeme davranış bozukluğunun biyokimyasal ve antropometrik ölçümlerle ilişkisinin belirlenmesi
A determination of the relations between orthorexia nervosa, abnormal eating attitudes and biochemical, anthropometric measurements in Type 1 diabetic adult patients on carbohydrate counting therapy
NESİL GÖREN ATALAY
- RFID entegreli envanter sayım robotu ile SLAM ve navigasyon yazılımlarının karşılaştırmalı performans analizi
Comparative performance analysis of SLAM and navigation software using an RFID integrated inventory counting robot
MEHMET ZEKİ SİLDİR
- Fetal hareket sayımı ve fetüs pozisyonu imajinasyonunun 24.-28. haftasındaki gebelerde prenatal bağlanma ve depresif belirtiler üzerine etkisi
Effectiveness of fetal movement counting and imagining fetus position to prenatal attachment and depressive symptoms in pregnant women in 24-28TH gestational week
LEYLA TÜRKER DEMİRKAN
- Semi - autonomous warehouse counting system
Yarı otonom depo sayım sistemi
GİZEM MELİKE CİDAL
Yüksek Lisans
İngilizce
2020
Yıldız Teknik ÜniversitesiMekatronik Mühendisliği Ana Bilim Dalı
Yükleniyor...
- Bir kablo donanım üretim sisteminde tam zamanında üretim uygulaması
Just in time application in a wiring assemblies and harneses production systems
MİNE AKSOY
Yüksek Lisans
Türkçe
1997
İstanbul Teknik ÜniversitesiEndüstri Mühendisliği Ana Bilim Dalı
Yükleniyor...