Building A Randomizer For Cargo Claims Predictions Using Computer Vision and Machine Learning
In this two-month Omdena Challenge, a global team of 50 AI changemakers collaborated to build a phone application that identifies safe delivery conditions of the boxes within a container.
The Problem and Background
A cargo claim is essentially a demand for monetary compensation in respect of financial loss sustained as a result of a breach of the contract of carriage or a failure by the carrier to fulfill certain extra-contractual obligations. However, the carrier will not be liable to compensate the claimant if the loss is caused by circumstances for which he is exonerated from liability by force of law or contract.
Reducing inefficiencies not only improves profitability but also has direct societal and environmental impacts.
To ensure the best delivery conditions and limit cargo claims, Omdena’s team worked on identifying the characteristics and conditions of boxes within a container, like safe storage temperature, stuffing patterns, and extent of damage from phone images.
The Challenge Outcomes
Using a phone application that we can point to a stack of boxes to randomly select which one to inspect.
Use available information around the characteristics of a product to determine the likelihood of damage and the settlement outcome.
Determine the impact of extended storage time and or increased temperature on temperature-sensitive products.
Scrape and existing case laws to highlight similarities and differences to cases under consideration.
Display a settlement % eg below:
Overcome object detection limitations in the case of having obstacles through the image, i.e. the box is cropped behind a column or object detection rectangles are misformed.
Examples of some difficulties in interpreting boxes’ information are in the images shown below.