One challenge in creating commercial solutions with supervised deep learning is acquiring large, customized labeled datasets. These datasets must often fit within commercial industries’ production times and budgets. There is still a need for target data that allows the model to learn specific characteristics of the target domain. The construction of customized datasets is relevant in imaging waste identification to improve the accuracy and efficiency of models, allowing adaptation to specific conditions. To address this challenge, we propose the EcoMind open-source tool. This tool facilitates waste labeling from user-provided images and empowers local dataset creation and collaborative building with users.
