----------------------- REVIEW 1 ---------------------
PAPER: 1305
TITLE: Fashionista: A Fashion-aware Graphical System for Exploring Visually Similar Items
AUTHORS: Ruining He, Chunbin Lin and Julian Mcauley

OVERALL EVALUATION: 0 (borderline paper)

----------- REVIEW -----------
This demo proposal is well-written though I would prefer that the background section came before the system description. The claimed contribution is clearly presented and seems appropriately novel: though could be located better within general image similarity work.

“rated similarly (by users) in terms of their appearance”: or purchased?

The proposal notes that the system needs user testing; a demo might be a good start for this future testing. The paper does provide several scenarios that the system can support: and the functional online system does enable these cases.



The references could have some more details: there are several missing volume and page numbers.


Presentational:
AJAX v Ajax: consistency


----------------------- REVIEW 2 ---------------------
PAPER: 1305
TITLE: Fashionista: A Fashion-aware Graphical System for Exploring Visually Similar Items
AUTHORS: Ruining He, Chunbin Lin and Julian Mcauley

OVERALL EVALUATION: 1 (weak accept)

----------- REVIEW -----------
The paper proposes Fashionista, a tool for aiding people to find appropriate clothes and shoes based on catalogs from e-commerce stores. It is based on deep learning algorithms and determines matches based on visual attributes.

The topic is very relevant and challenging, being addressed by state-of-the-art algorithms and techniques.

I have three criticisms to the paper:

1) the system does not seem to be available in the web, which would help the evaluation.

2) the paper may discuss in better detail what are the attributes considered in the learning
process.

3) as far as I understood, Fashionista does not provide personalized services. It would be interesting to discuss the challenges and potential solutions to this problem. As a personal assistant, providing personalized recommendations and evaluations seems to me as a key task.

Finally, the paper is well written and does not need significant revisions.


----------------------- REVIEW 3 ---------------------
PAPER: 1305
TITLE: Fashionista: A Fashion-aware Graphical System for Exploring Visually Similar Items
AUTHORS: Ruining He, Chunbin Lin and Julian Mcauley

OVERALL EVALUATION: 3 (strong accept)

----------- REVIEW -----------
The proposal presents an interactive product recommendation system built on top of a deep learning backend that groups products based on visual similarity. The work nicely combines theoretical concepts to a potentially useful application. The visual interface looks feature-rich yet easy-to-use. Although presented as a fashion recommendation system, the concept would apply to products from other domains too. Overall, an interesting work suitable for the demo session.

Minor suggestion: the fashion trend exhibitor can be enhanced to a general-purpose container that switches to different analytics on-demand.