lukas/bachelor-thesis
1
0
Fork 0
Code Issues Pull Requests Actions Packages Releases Activity

fix stuff from prof
Some checks failed
Build Typst document / build_typst_documents (push) Failing after 1m39s
Details

Browse Source
This commit is contained in:
lukas-heiligenbrunner
2025-02-04 20:07:59 +01:00
parent 7681b4afce
commit c8ac00571b
6 changed files with 27 additions and 27 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
implementation.typ
View File

@ -4,7 +4,7 @@
#import "@preview/subpar:0.1.1"
= Implementation <sectionimplementation>
The three methods described (ResNet50, CAML, P>M>F) were implemented in a Jupyter notebook and compared to each other.
The three methods described (ResNet50@resnet, CAML@caml_paper, P>M>F@pmfpaper) were implemented in a Jupyter notebook and compared to each other.
== Experiments <experiments>
For all of the three methods we test the following use-cases:
@ -29,7 +29,7 @@ The rest of the images was used to test the model and measure the accuracy.
== ResNet50 <resnet50impl>
=== Approach
The simplest approach is to use a pretrained ResNet50 model as a feature extractor.
The simplest approach is to use a pretrained ResNet50@resnet model as a feature extractor.
From both the support and query set the features are extracted to get a downprojected representation of the images.
After downprojection the support set embeddings are compared to the query set embeddings.
To predict the class of a query, the class with the smallest distance to the support embedding is chosen.
@ -136,7 +136,7 @@ but this is expected as the cable class consists of 8 faulty classes.
== P>M>F
=== Approach
For P>M>F, I used the pretrained model weights from the original paper.
For P>M>F@pmfpaper, I used the pretrained model weights from the original paper.
As backbone feature extractor a DINO model is used, which is pretrained by facebook.
This is a vision transformer with a patch size of 16 and 12 attention heads learned in a self-supervised fashion.
This feature extractor was meta-trained with 10 public image dasets #footnote[ImageNet-1k, Omniglot, FGVC-
@ -182,7 +182,7 @@ So it is clearly a bad idea to add more good shots to the support set.
== CAML
=== Approach
For the CAML implementation I used the pretrained model weights from the original paper.
For the CAML@caml_paper implementation I used the pretrained model weights from the original paper.
The non-causal sequence model (transformer) is pretrained with every class having the same number of shots.
This brings the limitation that it can only process default few-shot learning tasks in the n-way k-shots fashion.
Since it expects the input sequence to be distributed with the same number of shots per class.