GSOC Log jiyeon's log

[Week 10] time for perfecting the latency estimator (3)

Get real model’s params for getting regressor score Model used : densenet, MobileNet from tflite model zoo, and hair segmentation, face detection from mediapipe github total : 245 samples But it seems worse… I think it would be better to include some of benchamrks of the real model in the train set. Not all, but adding about 20% would ... Read more

[Week 9] time for perfecting the latency estimator (2)

Selection of Regression models? Decision Tree and RandomForest Tree model change dramatically depending on the value of the random state. XGBoost has always shown consistent good performance. First version of Estimator Output : csv file or command line tool? Read more

[Week 8] time for perfecting the latency estimator

Two MOST important things to do 1. Database training one with manually generated data - NOW fixed! The amount of benchmarked models is 908 testing one from the real model Last week, I mentioned that I was able to print parametors of convolution of the real model In this week, I tried to apply the code to the real model and now I ca... Read more

[Week 7] Adjust database with an updated set of convolutions

Things to fix for database building Clean up the github repo link : First of all, I have only uploaded files that I created/modified. I’m not sure it is okay. What files should be included specifically? Metric There are 3 main metrics for model evaluation in regression. 1. R Square (-> our current... Read more

[Week 6] Analyze what was missed in the database

Add columns on database I got more database samples by adding padding and stride columns and putting various input shape. Initial version of regression model In this week, I tried to make simple regression model using scikit-learn library in python. # data preprocessing import os import re import numpy as np import pandas as pd import matp... Read more