What role Computer Vision + Deep Learning can play in a Warehouse?

Off late there is lot of disruption happening in Warehouse industry. Growing ecommerce adoption globally, there is a lot of pressure on Warehouses and storage houses to enhance their throughput. 

There are lot of innovative use cases that are possible with Computer Vision & Deep Learning, possibilities are endless.

At SandLogic, we helped one of our customer who have 40+ warehouses across US by developing a very interesting and challenging use case using AI. You can find more details of the same @ https://www.sandlogic.com/deep-connect/pallet-scanner/

They have on an average 500+ pallets inbound and 500+ outbound per warehouse and updating the inventory was taking lot of time, we have used Computer Vision to automate the process of classification of pallet and also, count the number of boxes on the pallet and update the WMS (Warehouse Management systems) accordingly. This use case significantly reduce the manual efforts and also, reduces human error.

Challenges were getting the box count when there is a polythene wrapped around the pallet and there is a reflection due to light and also, completing the whole process by the time pallet completes it movement from docking area to putaway, yes i.e. in few seconds.... :)

We have created

1. Custom model on top of ResNet architecture
2. Dataset from the Scratch
3. Annotations using our home grown annotation tool AuVi.io

Pasting here a sample illustrative image that is fed to our AI Model, helps to guage the complexity involved.