The dataset viewer is not available because its heuristics could not detect any supported data files. You can try uploading some data files, or configuring the data files location manually.
Dataset Card for innv-cosmetics-dataset-for-retail
The dataset includes images / frames annotated with detailed labels for individual cosmetic products and planograms (shelf layouts).
This dataset comprises a collection of annotated frames extracted from videos recorded in retail cosmetic stores. It was initially intended as a first test for a larger project, requiring iterative feedback and additional quality reviews to meet evolving client expectations. Despite these efforts, the collaboration was marked by challenges, including unclear evaluation criteria and a lack of timely communication from the client. The rights to the original video content belong to their respective owners, and Innovatiana does not claim any rights to these images or videos. Innovatiana's role was to annotate and structure the dataset, providing a valuable resource for AI developers working on retail and inventory-related AI use cases.
(!) Please note that this dataset shouldn't be considered as fully comprehensive: as per its purpose (first test for further adjustment), some minor flaws or lack of annotation remain (refer to # Dataset_Structure for more details).
Dataset Story
The story of this dataset mirrors the journey of countless data annotators worldwide—dedicated professionals who work tirelessly to produce high-quality datasets, often under challenging conditions. Unfortunately, their work is sometimes undervalued or dismissed due to vague evaluation criteria and a lack of accountability from clients. This dataset reflects not only their diligence but also the systemic challenges that plague the data annotation industry.
In this particular case, Innovatiana completed the annotation work in good faith, delivering a 1st dataset of good quality. However, disputes arose based on unclear criteria, compounded by the client’s prolonged lack of communication. Despite going silent, the client proceeded to use the dataset to train their AI model, effectively benefiting from the work without honoring their financial commitments.
The good news is that the agreements governing this project allow us to release the dataset to the Open Source community. By doing so, we not only honor the work of our annotators but also turn this experience into a meaningful contribution to the AI field. This release stands as a testament to transparency, collaboration, and our commitment to fair practices, ensuring that this dataset serves as a valuable resource for AI developers.
(!) It is important to clarify that the client for this dataset was NOT affiliated with any of the brands or stores depicted in the images. HOWEVER, we believe that these brands, as potential end beneficiaries of AI solutions developed by our client, should be made aware of how some suppliers procure their training data—often through unethical practices such as unauthorized usage or stealing data.
Dataset Description
Content: The dataset features:
- 4,820 annotated frames
- Approximately 245,000 labels
- Complex annotations with up to 500+ labels per frame in some cases
Sources: Frames were extracted from in-store video footage recorded in cosmetic retail environments.
Format: Raw images and their respective annotations, in XML format are available. Dataset was developed using CVAT and organized into subsets to optimize performance during extraction and analysis.
Annotation tool used: CVAT
Annotations: Annotations include bounding boxes / polygons for individual products and shelf layouts. For projects requiring additional metadata, feel free to contact us.
Dataset Usage
This dataset is intended to support AI use cases in retail and inventory management, including but not limited to:
- Inventory Tracking: Automating the recognition and tracking of products on shelves.
- Shelf Layout Optimization: Analyzing planograms to ensure compliance with retail standards.
- Object Detection: Developing models to identify and classify cosmetic products.
- Stock Deficiency Detection: Identifying empty shelf spaces or missing products in real-time.
Dataset Structure
innovatiana-cosmetics-dataset-for-retail > Refer to innv-cosmetics-dataset-for-retail.pdf for details
(!) Please be aware that not all subsets have passed our Data Quality review. Some remain incomplete (with missing annotations), while others were utilized for training purposes and may include random or inconsistent data/metadata, making them unsuitable for direct use in AI training.
(!!) For a detailed overview of the subsets that are ready to use, please refer to the innv-cosmetics-dataset-for-retail.pdf. Regardless, we strongly recommend refining the data to align with your specific use case, as this will save significant time and effort in the long run.
Licence
This dataset is intended for educational and research purposes only. All product images are the property of their respective copyright holders. Users must ensure they comply with applicable laws and regulations when using this dataset. Innovatiana does not claim any rights to these images.
Contact
For any questions or issues, please open an issue on this Hugging Face repository or contact the maintainer at info [at] innovatiana [dot] com
Disclaimer
This dataset is not affiliated with or endorsed by the brands visible in the dataset. It is created solely for academic and research purposes. Please reach out to info [at] innovatiana [dot] com if you are looking to enrich this dataset or to create a similar one for a commercial usage.
- Downloads last month
- 40