Research Article
Hybrid Approach for Shelf Monitoring and Planogram Compliance (Hyb-SMPC) in Retails Using Deep Learning and Computer Vision
| Input: D: Labeled dataset D = {(), (), (), …, ()} with p images | | B = {, , , …, } with q classes | | M = {, , …, with j models | | IM = input images | | Output: Trained models | | = labels of Classes for the SKUs included in input images | | Start: | | = Split (D, p ∗ 80) | | = Split (D, (p-(p ∗ 80))) | | //Step 1—Training of models with labeled data | | for n = 1 to j: | | for every epoch: | | for every () in: | | = Train () | | end | | end | | end for | | //Step 2—Testing models | | for k = 1 to j: | | for every () in : | | Prediction = () | | end | | end for | | //Step 3—Detecting SKUs in input imagebˆ = TM () | | Output: Processed images (PI) | | //Step 4—Sorting SKUs and Racks | | PPI = Sorting (PI) | | //Step 5—Generating Planogram from JSON object and comparing post processed image with Planogram layout | | JO = contour (Pg) | | foreach in DCompare (, JO) | | End Algorithm |
|