Project 14. Android based Object Detection and Classification: Modeling a Child's Learning of What's Hot and Cold

Should this project be considered for the Best Project award? No.
Should this project be considered for the top 3 project awards? No.

On a scale of 1 to 10, how would you rate the overall organization/clarity?
5/10

On a scale of 1 to 10, how would you rate the overall project idea? 
6/10

On a scale of 1 to 10, how would you rate the overall research
contribution of the project idea, methodology and/or results?
6/10

Overall, the project communicates what was done adequately. What is not clear is why it was done. The report discusses
developmental learning, but performs object detection in the style of run-off-the-mill computer vision. The results are not reported very clearly -- it would be nice to see the overall accuracy of the recognition as a function of the number of training examples.

Overall, the spelling of words in the report is correct, but their order is not always correct. There are also some problems with capitalization, punctuation, and grammar (including an ungrammatical sentence in the abstract). 

It is somewhat hard to read text that is at times not broken up into paragraphs. There is also a number of interesting mis-wordings, including the one in the abstract (unless the system was indeed intended to tell hot from code based on previous experience, which would certainly be interesting). In particular, the word sudo,  which also stands for the command for "execute as root" in Unix, often replaced "pseudo". Another interesting typo is "verse" in place of "versus". I can only assume that these errors result from a misguided application of an automated spell checker.

What is missing from the report is situating the system within the overall field of Computer Vision. The SOM as a learning algorithm is somewhat dated and, thus, requires extensive justification in the broad context of the related work. This is a time-consuming task, especially given the amount of publications in Computer Vision. One state of the art algorithm for object class recognition using vision is described in [1].

The project did produce a working system and does meet the course requirements. I am sure that implementing the system was a lot of fun. It was also nice to see the Use Case analysis for the UI.
It was also very nice to see that the code was made open source.

[1] Fergus et al. Object class recognition by unsupervised scale-invariant learning.