February 2024 Summaries

This week's FiftyOne tips and tricks blog covers various questions and answers from Slack, GitHub, Stack Overflow, and Reddit. Topics include using dataset.export to specify arbitrary file paths, clearing selected samples in the FiftyOne App programmatically, adding custom attributes, working with very long video samples, and filtering all unlabeled samples programmatically. The blog also provides solutions for each question and directs readers to relevant sections of the FiftyOne Docs for further information.

Voxel51 has released FiftyOne 0.23.5 and FiftyOne Teams 1.5.6, offering numerous enhancements and fixes for computer vision workflows. These updates include support for monocular depth estimation transformers from the Hugging Face library, subcounts in search results, and improved dataset search user experience. The company encourages users to join their open-source community through various channels such as Slack, Meetups, and GitHub. They also provide an industry spotlight blog series that highlights how computer vision is being utilized across different sectors, with the latest focusing on retail applications.

This week's FiftyOne tips and tricks blog covers various topics such as computing distinct values of a field in a collection, finding dissimilar images using the "similarity" feature, exporting labels into a YOLO format, cleaning up previous evaluation runs, and reloading datasets in the FiftyOne App. The blog also provides code snippets to help users implement these features effectively.

The February AI, Machine Learning and Data Science Meetup covered topics such as fair computer vision, food waste classification with AI, image geo-localization from visual data, and the next generation of video understanding. Key speakers included Chenliang Xu, Luka Posilović, Safwan Wshah, and James Le. The event also featured a Q&A session for each presentation. Upcoming Meetups will take place on March 21st at various locations worldwide.

Model Explainability is a crucial field in machine learning as it helps understand how models work and identify hidden biases or misclassifications. Class Activation Mapping (CAM) methods, such as GradCam, are algorithms that produce heatmaps on images to show where the model focuses for specific classes. These methods can be used with FiftyOne to gain a better understanding of models' predictions and spot potential biases. By leveraging CAM methods, we can improve our models' performance and ensure they focus on relevant features rather than environmental factors.

In this monthly blog series, we celebrate recent milestones in the FiftyOne community. Kitro, a Swiss company using FiftyOne for sustainable change, is highlighted as an example of how the open source computer vision toolkit helps organizations develop machine learning models and curate datasets. Additionally, new features such as Hugging Face integration and SuperGradients integration are introduced. The blog also provides updates on product releases, community engagement, and upcoming events related to computer vision technology.

The retail industry is leveraging computer vision and artificial intelligence to drive innovation. Key applications include inventory management, autonomous checkouts, virtual try-ons, customer behavior analysis, and product recommendations. Computer vision technologies help retailers optimize store layouts, enhance product placement, tailor marketing strategies, reduce food waste, and improve the overall shopping experience. Companies such as Trigo, Trax Retail, Link Retail, Dayta AI, and RetailNext are at the forefront of implementing these technologies in the retail sector.