Week #4 #

External Feedback #

Since our project is set by the client, we use the feedback from them to adjust and refine our project. Due to having a clear goal set, we possibly don’t need to worry about the feedback as much as other teams, however it still helps while we discuss necessary features, usable techologies or general project functionality. The most recent feedback we got was from our prototype presentation, where we received the following responses (translated from Russian):

• “The guys did a great job, they made a very cool UI. Such a UI and working prototype is very difficult to build in such a short time, well, here in our company, for example.”
• “Nice images, quickly generated, it looks very cool.”
• “We would like to continue to explore this solution. May be we can use it somehow.”

Overall, the clients are pleased with our current progress and are interested in our team’s development.

Testing #

As our user base is limited to our clients, the testing is done mostly by us. That’s why our team carefully checks every new part, since the clients cannot allocate time for the full testing.
The most effective approach that we found is to make members of our team cross-check other parts of the project that they have little knowledge of, as they are more likely to point out the inconsistencies and poor design choices.

Iteration #

Since our team has a clearly formulated goal set by the customer, we do not deviate from the original vision, and the project progresses completely in the right direction. As such, it is in our best interest to continue adding the intended features and refining completed ones in our prototype.

Weekly Progress Report: #

Challenges & Solutions #

• Hard dataset preprocessing
  Solution: image segmentation and manual processing
• Sudden problems in the prompt generation after upgrades
  Solution: trial and error :)

Conclusions & Next Steps #

Our team made further progress on the project, such as adding style selection, refining the machine learning pipeline, connecting the frontend and backend and e.t.c. With all of this in mind, our next steps are:

• fixing bugs
• adding the negative promts
• rename labels for better transfer to Yandex GPT
• add saving pictures and registration