1.新年を楽しく過ごす
2.論文を読みました ：
Explanation-based Weakly-supervised Learning
of Visual Relations with Graph Networks（https://www.paperswithcode.com/paper/explanation-based-weakly-supervised-learning）；
Visual Relationship Detection with Deep Structural Ranking（https://www.paperswithcode.com/paper/visual-relationship-detection-with-deep）；
3.宿題を済ます
進捗
・NTTデータ座談会
・システム開発会社本選考 筆記試験受験
予定
・英論修正
・NTTデータOB訪問
・アドソル日進 本選考筆記試験受験
TOTOのwebテスト受験
TDKのwebテスト受験
I have been in KIT for one month.Thanks for every one’s help to me.And i think i am customed to the life there.
For the first week(12.1-12.6),i have many entrance procedure to do.
Second week and third week(12.7-12.20),take classes and finish the GCN classes in BAIDU.
The third week(12.21-12.27).take classes and review the object detection classes in BAIDU.
Many times are spent on finishing the homework of the classes, and i also want to learn Japanese.There are much things for me to do.
In the following time, after i finish the detection task in BAIDU by paddle,i will try to make a demo to Visual-Relation-Detection based on Cascade-RCNN by Paddle platform.
進捗 幾つかの画像についてプログラムを走らせ、プログラムの欠陥の修理を行った。
課題 現在プログラムが一回走るごとに一枚の検出のみなので、時間があれば複数同時の識別に対応させたい。
日鉄ソリューションズ本選考ES作成中
止まれ、侵入禁止、左折禁止の標識を識別するプログラムの統合を行なった。
欠損率を求めるプログラムを試作した。
I have been in KIT for one month.Thanks for every one’s help to me.And i think i am customed to the life there.
For the first week(12.1-12.6),i have many entrance procedure to do.
Second week and third week(12.7-12.20),take classes and finish the GCN classes in BAIDU.
The third week(12.21-12.27).take classes and review the object detection classes in BAIDU.
Many times are spent on finishing the homework of the classes, and i also want to learn Japanese.There are much things for me to do.
In the following time, after i finish the detection task in BAIDU by paddle,i will try to make a demo to Visual-Relation-Detection based on Cascade-RCNN by Paddle platform.
I modified my AROB thesis, and submit the thesis to the AROB last week. From the 3D object detection survey, I learned a lot about the 3D detection works, and find the frameworks using cameras would be better than those using lidars.
Next I would try … Continue Reading ››
転移学習を用いたプログラムの作成
過学習を起こしており、精度が低いためそれらの問題点について解決策を模索している