Real-time Data Engineering Specialist Needed
Budget: £20 – £250 GBP
I have a javascript (NextJS) project in which I want to take a photo of a "Scorecard" and recognise the digit in each cell of the grid. The grid will be 10x3 cells, but not all will be used, (although I do know in advance which ones are expected)
I already have working code to take a photo, identify the grid corners, and apply a homography transform to (approximately) straighten up the image into a 500x150px image. I have attached a few examples. I now need to extract the contents of each cell. I will then use tensorflow to provide digit recognition for each cell. I already have the digit recognition model for TF.
***This project is to take an image like the attached, identify the grid lines, and then extract the contents of each cell, ensuring the gridlines are omitted.*** There is no need to further process the cells or perform digit recognition.
You could do this either by erasing the gridlines and then simply slicing the grid (10x3), or by finding the maximum bounding box which fits inside the gridlines for each cell, or perhaps by "growing" a region from the center of each cell. My aim is to isolate the written character from the grid cells by whatever method works well, and fast.
I am NOT currently importing openCV, because of the large file size. I would prefer not to use openCV if it can be avoided.
I am open to suggestions for alternative approaches. Budget negotiable, but I think this should be straightforward for an experienced Computer Vision practitioner.
I already have working code to take a photo, identify the grid corners, and apply a homography transform to (approximately) straighten up the image into a 500x150px image. I have attached a few examples. I now need to extract the contents of each cell. I will then use tensorflow to provide digit recognition for each cell. I already have the digit recognition model for TF.
***This project is to take an image like the attached, identify the grid lines, and then extract the contents of each cell, ensuring the gridlines are omitted.*** There is no need to further process the cells or perform digit recognition.
You could do this either by erasing the gridlines and then simply slicing the grid (10x3), or by finding the maximum bounding box which fits inside the gridlines for each cell, or perhaps by "growing" a region from the center of each cell. My aim is to isolate the written character from the grid cells by whatever method works well, and fast.
I am NOT currently importing openCV, because of the large file size. I would prefer not to use openCV if it can be avoided.
I am open to suggestions for alternative approaches. Budget negotiable, but I think this should be straightforward for an experienced Computer Vision practitioner.