Hi, I am conducting research for my final project. I am designing and building a machine learning model to classify oil palm fruit into three categories, using an ESP32-CAM microcontroller with an embedded SVM model. The goal of this project is to address existing issues in the oil palm harvesting process—specifically, errors and harvest failures that compromise fruit quality or yield. Manual harvesting relies on individual visual judgment, which varies from person to person; the device I am building, powered by ML, aims to solve this problem. I specifically want to expand and improve my research by using Edge Impulse more effectively.
However, I am encountering an obstacle with the SVM classification feature under the “Classifier” section within “Complete architecture / classic ML.” It appears that access to this feature is restricted to enterprise accounts. How can I implement the SVM model in this case? Could someone please help me? I have repeatedly registered for a student account to gain access and signed up for the trial, but I haven’t received any reply or further information. I am unsure if I selected the wrong trial option or what the issue might be; could you please explain this to me?
I’ll check on the entitlements for students, but in the meantime, can you tell me why you want to use an SVM? Do you have a feature extractor in mind, SVM will perform poorly on raw pixels. I assume this is a visual classification project? I would suggest starting with EfficientNet, or you could also try Yolo Pro (even though it will give you bounding boxes you can just disregard them and use it like a classifier)
My current project involves classifying fruit types based on color, using the SVM algorithm. Why SVM? Because, according to the references I’ve consulted, this algorithm performs quite well in classification tasks. Furthermore, since other studies have already utilized many different methods and algorithms, I decided to apply a new classification algorithm SVM in this research project.
A brief explanation of my research project: I’m focusing on evaluating the implementation results and the accuracy of fruit ripeness classification using the SVM algorithm. I’ll also analyze the results in the “Confusion Matrix” and assess the computational performance of the ESP32-CAM microcontroller itself. For this reason, I’m using the Edge Impuls platform for my data training process, and after the training process is complete, I will deploy it to the ESP32-CAM.
Actually, my process is almost complete; I’m just facing an issue with accessing or using the company’s labels when applying SVM to select a complete architecture/ML. If I can implement and access that, it will be finished.
I’ve also tried to sign up for a trial account for the company label, but I haven’t received a response yet, and the process is a bit complicated, so I’m confused. As far as I know, the trial sign-up process usually just involves filling out a form or entering some information and a credit card, but on this platform it’s a little different. Is there a solution I can use to ensure this project runs smoothly and gets completed? Thank you.