Update README.md
Browse files
README.md
CHANGED
@@ -1,3 +1,34 @@
|
|
1 |
---
|
2 |
license: apache-2.0
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
---
|
2 |
license: apache-2.0
|
3 |
+
tags:
|
4 |
+
- en
|
5 |
+
- image classification
|
6 |
+
- computer vision
|
7 |
+
- image recognition
|
8 |
+
model-index:
|
9 |
+
- name: ohbugger2k by flobbit
|
10 |
+
results:
|
11 |
+
- task:
|
12 |
+
name: insect image classification
|
13 |
+
type: image-classification
|
14 |
+
metrics:
|
15 |
+
- name: accuracy
|
16 |
+
type: acc
|
17 |
+
num_train_epochs: 7
|
18 |
+
learning_rate: 0.002
|
19 |
+
value: 46.0
|
20 |
---
|
21 |
+
|
22 |
+
# Oh! Bugger! 2k
|
23 |
+
|
24 |
+
## Model description
|
25 |
+
|
26 |
+
The model is used to classify insect images into one of the 2000 species/classes. `resnet18` was used for training.
|
27 |
+
|
28 |
+
## Intended uses & limitations
|
29 |
+
The model was trained on 130133 insect images spread over 2000 species with a minimum of 25 pics in a class. Some classes were trained on too few images. The training pics were not screened for quality. For example, giving the model a picture of a human finger will most likely return an insect species that had finger in a training pic. Or a pic of a bug on the siding of a house will likely return that type of match.
|
30 |
+
There are likely other biases in the training data.
|
31 |
+
|
32 |
+
## Training and evaluation data
|
33 |
+
|
34 |
+
The images used in training were scrapped from the internet.
|