MUMBAI, India, May 29 -- Intellectual Property India has published a patent application (202641061830 A) filed by Vignan's Institute Of Information Technology, Visakhapatnam, Andhra Pradesh, on May 15, for 'fruit freshness detection using ensemble deep learning and grad-cam based explainability.'

Inventor(s) include Ms. Ch. Usha; Malla Sravan Kumar; Pentapalli Kushal; Beesetty Likhita; Maddala Rakshitha; and Potnuru Chandra Sekhar.

The application for the patent was published on May 29, under issue no. 22/2026.

According to the abstract released by the Intellectual Property India: "An automated system for checking if fruits are fresh or not is proposed. It utilizes deep learning and explainable AI. The objective is to classify fruits into Fresh or Rotten categories for 13 different types of fruits. Currently, most outlets depend on human labor to inspect every fruit, but this is not efficient, and human judgment is not cutting it. So what's new here? We developed a single pipeline to use the combined output of three different convolutional neural networks: ResNet50, EfficientNetB0, MobileNetV3. The output of each model is combined by the simple mechanism of averaging our output, making the output stable and accurate for us. The entire process is automated: image preparation, prediction, display of results. To aide users in believing what is really seen by the AI, we make use of Grad-CAM. It acts, in a sense, as a kind of heatmap that shows where, specifically, certain fruit images impacted the decisions made by the model. The total number of collected images was 9,974, consisting of fresh and rotten images in 13 different fruit categories. With regard to the classification of the models, EfficientNetB0 has the" highest accuracy," 93.7%, compared to ResNet50, which has 92.3%, and then comes MobileNetV3 with 90.8Those Grad-CAM maps? Well, they show that the models aren't just guessing. They are picking areas that truly are problematic, such as a fungal growth, unusual textures, and brown areas, particularly for identifying rotten fruits. Needless to say, this is an excellent application for a fruit quality checker."

Disclaimer: Curated by HT Syndication.