Update README.md
Browse files
README.md
CHANGED
@@ -1,3 +1,25 @@
|
|
1 |
---
|
2 |
license: apache-2.0
|
|
|
|
|
|
|
|
|
|
|
|
|
3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
---
|
2 |
license: apache-2.0
|
3 |
+
datasets:
|
4 |
+
- detection-datasets/coco
|
5 |
+
language:
|
6 |
+
- en
|
7 |
+
library_name: keras
|
8 |
+
framework_version: 2.13
|
9 |
---
|
10 |
+
|
11 |
+
# Model card for keras_nanodet_plus_x1.5_416x416
|
12 |
+
|
13 |
+
A keras implementation of Nanodet-plus-x1.5 object-detection model, pretrained on COCO dataset.
|
14 |
+
|
15 |
+
The implementation is based on [RangiLyu/nanodet](https://github.com/RangiLyu/nanodet), with input resolution of 416x416. The quantized models were generated by [Model Compression Toolkit](https://github.com/sony/model_optimization).
|
16 |
+
|
17 |
+
|
18 |
+
## Model Quantization
|
19 |
+
- **Avaiable quantizated models**
|
20 |
+
- PTQ_W8A8
|
21 |
+
- **Accuracy**
|
22 |
+
- Float: 34.1
|
23 |
+
- PTQ_W8A8: 33.2
|
24 |
+
- **Size estimation:**
|
25 |
+
- PTQ W8A8: 2.44MB
|