Fresko Pisces: Fish Freshness Identification Using Deep Learning
| Autor: | Anju Pratap, Chinju Philip, Jithu George Velloor, Anandhu Suresh, Arathi Vinayachandran |
|---|---|
| Rok vydání: | 2021 |
| Předmět: |
Ground truth
Contextual image classification Computer science business.industry Deep learning Machine learning computer.software_genre Convolutional neural network Field (computer science) Identification (information) Artificial intelligence Transfer of learning business computer Feature detection (computer vision) |
| Zdroj: | Innovative Data Communication Technologies and Application ISBN: 9789811596506 |
| Popis: | Fish freshness identification plays a prominent role in fishery industry applications. Recently, serious malpractices are observed in fishery industries, which is highly becoming as a socioeconomic issue. The recent development of convolutional neural networks has accomplished great outcomes in the field of image classification. This paper focuses on novel method of dealing with the advancement of fish freshness identification utilizing convolutional neural network. VGG-16 network through transfer learning is used as the classification algorithm. This will predict the percentage of freshness, remaining shelf life of a fish on giving its eye image, gill image, and skin discoloration like features as input. The data set was prepared by collecting more than 6000 real samples from various fish markets and shops of Kerala using cameras during a time period of 3 to 4 months. After data set collection, pre-processing was done in order to select the most suitable images and the model. The Final conclusion about freshness is formed after analyzing the condition of the eyes and the gills and some manual method of feature detection. The specialty of the strategy exists in the limit of the strategies to viably acclimate to new examples which could be an after effect of the comprehensive propriety of the specified parameters. The system shows up the exceedingly exact outcomes when contrasted with the ground truth. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|