# Alternatively, if you do not have permissions or prefer
# not to install the COCO API into global site-packages
python2 setup.py install --user
**data reader introduction:**
* Data reader is defined in `reader.py`.
...
...
@@ -103,18 +118,7 @@ Finetuning is to finetune model weights in a specific task by loading pretrained
## Evaluation
Evaluation is to evaluate the performance of a trained model. This sample provides `eval_coco_map.py` which uses a COCO-specific mAP metric defined by [COCO committee](http://cocodataset.org/#detections-eval). To use `eval_coco_map.py` , [cocoapi](https://github.com/cocodataset/cocoapi) is needed. Install the cocoapi:
# Alternatively, if you do not have permissions or prefer
# not to install the COCO API into global site-packages
python2 setup.py install --user
Evaluation is to evaluate the performance of a trained model. This sample provides `eval_coco_map.py` which uses a COCO-specific mAP metric defined by [COCO committee](http://cocodataset.org/#detections-eval).
`eval_coco_map.py` is the main executor for evalution, one can start evalution step by:
...
...
@@ -136,7 +140,7 @@ Faster RCNN mAP
| Detectron | 8 | 180000 | 0.315 |
| Fluid minibatch padding | 8 | 180000 | 0.314 |
| Fluid all padding | 8 | 180000 | 0.308 |
| Fluid no padding |6 | 240000 | 0.317 |
| Fluid no padding |8 | 180000 | 0.316 |
* Fluid all padding: Each image padding to 1333\*1333.
* Fluid minibatch padding: Images in one batch padding to the same size. This method is same as detectron.