AI Bouquet Recommendation & Identification
Budget: €250 – €750 EUR
My goal is to turn the product photos from our online flower shop into a smart assistant that can do two things:
• look at any combination of individual SKU images and propose a visually coherent bouquet image, and
• look at a customer-supplied bouquet photo and tell us which exact SKUs they should buy to recreate it.
I would like this built with Google Cloud Vision AI for detection/classification and Roboflow for dataset management, augmentation, and model hosting. We already have a partial library of labeled flower images, but a portion of the catalogue still needs annotation, so guidance on the most efficient labelling workflow inside Roboflow will be appreciated.
Key expectations
– Configure the end-to-end pipeline: dataset upload, labelling strategy for missing SKUs, model training, evaluation, and deployment.
– Produce an inference routine that, given selected SKU photos, generates an image-only bouquet recommendation (no text list required at this stage).
– Produce a reverse-lookup routine that ingests a bouquet photo and outputs the detected SKUs with confidence scores.
– Deliver clean, reproducible code or notebooks plus any Roboflow/Cloud Vision configuration files so we can retrain as our catalogue grows.
– Final demo should run on a small sample set from our shop and achieve reliable SKU detection and visually appealing bouquet mock-ups.
If you have proven experience combining Roboflow with Google Cloud Vision (or similar CV frameworks) for product recommendation or visual search, I’d love to collaborate and get this into production quickly.
• look at any combination of individual SKU images and propose a visually coherent bouquet image, and
• look at a customer-supplied bouquet photo and tell us which exact SKUs they should buy to recreate it.
I would like this built with Google Cloud Vision AI for detection/classification and Roboflow for dataset management, augmentation, and model hosting. We already have a partial library of labeled flower images, but a portion of the catalogue still needs annotation, so guidance on the most efficient labelling workflow inside Roboflow will be appreciated.
Key expectations
– Configure the end-to-end pipeline: dataset upload, labelling strategy for missing SKUs, model training, evaluation, and deployment.
– Produce an inference routine that, given selected SKU photos, generates an image-only bouquet recommendation (no text list required at this stage).
– Produce a reverse-lookup routine that ingests a bouquet photo and outputs the detected SKUs with confidence scores.
– Deliver clean, reproducible code or notebooks plus any Roboflow/Cloud Vision configuration files so we can retrain as our catalogue grows.
– Final demo should run on a small sample set from our shop and achieve reliable SKU detection and visually appealing bouquet mock-ups.
If you have proven experience combining Roboflow with Google Cloud Vision (or similar CV frameworks) for product recommendation or visual search, I’d love to collaborate and get this into production quickly.