Sprint 5 Summary:ui interface Update, Azure-side deployment and user feedback analysis 12/28/2015

The main task of this sprint is to update the UI interface of the mobile app, and to learn and program the deployment of the Azure client accordingly. The user feedback for Alpha release is also analyzed to determine the focus of the next work.

Work Progress:

1. UI update and improvement work mainly by the Zhao Yang, mainly in the app in the automatic labeling and navigation layout of the corresponding improvements. Most of these are different from the interface in Sprint2, and there are some improvements to the overall app performance. The corresponding improvement feature are as follows:

1). Personal activity Category Browse: It classifies the user's photos according to the content, time, location information. This allows users to browse through photos based on different types of activities, making it easy for users to experience.

2). Search recommendation: For the user to enter the words to predict, and recommend the corresponding image thumbnail, so that users can not spell the exact words of the exact time of the search.

3). Voice Search: Using the combination of Oxford Speech API and Stanford NLP API, the function of speech fuzzy search has been realized successfully. Users can directly say a sentence, using the Oxford Speech API to achieve voice-to-text conversion, and then use the Stanford NLP API to extract the keywords in the text to use as the last search keyword.

4). Automatic Label Generation: Use the popular deep neural network model CNN to process the image and generate the corresponding label accordingly. At the same time, the image of the existing GPS, shooting time and other information also carry out the corresponding reservation and layout processing to facilitate user browsing.

2. At the same time the background processing was optimized, mainly to have the NLP resful from the eclipse environment to the Tom Cat, which is responsible for the part of Minlone.

3. The corresponding code integration work is focused on the optimizations of some threading mechanisms, while some attempts have been made to transfer the code to the iOS system, which is primarily the responsibility of whisk and yandong.

4. In terms of Azure server deployment, the Azure architecture was first studied and a corresponding attempt was made in the area of deployment, which is primarily the responsibility of the building and Yandong.

Azure-Side deployment Scenario Analysis:

The azure-side deployment is currently in a trial phase, and after receiving the appropriate user feedback, we use the appropriate sub-decision to make the azure-side product an intermediate transition product, and the final effort will turn to a new attempt at the offline version of the app.

User Feedback Results Analysis:

A corresponding user has the following comments:

1. For this mobile app, I sometimes upload photos to receive the speed limit, unable to quickly get results. And the overall return of the picture results is good, but some are not particularly ideal. User Activity classification This special tastes, for me to browse the corresponding photos save a lot of trouble!

2. Overall performance is good, the details can be. But sometimes return to some inexplicable results, compared to no words ~

For the above two user reviews analysis can be made as follows some of the areas that can be improved:

1. The use of the popular CNN model in the actual use of the application can not be simple, perhaps through the corresponding threshold or simple operation implies a user experience improvement.

2. The online version is limited by other factors such as the speed of the network, and the implementation is also quite complex, in the long-term view, this is a good choice. But for the time being to achieve the same performance in the short term, it is not easy to complete the CNN iOS model architecture. The offline version may be the ultimate goal, but it will never be the final result of this course. The road to software development is a heavy burden, and may only be extremes meet.

Sprint 5 Summary:ui interface Update, Azure-side deployment and user feedback analysis 12/28/2015