@@ -22,7 +22,7 @@ Here are the slides (in german) for the two presentations we gave during the sem
## Dataset
We used the [**COCO**](http://cocodataset.org/#explore) Dataset [[PDF](https://arxiv.org/pdf/1405.0312.pdf)].
We used the [**COCO**](http://cocodataset.org/#explore)2017 Dataset [[PDF](https://arxiv.org/pdf/1405.0312.pdf)].
It provides:
* ~118000 pictures, manually annotated with contained objects
* 80 object categories
@@ -337,6 +337,7 @@ multilabel:
| Bleu-4 | 0.141 | 0.140 | 0.309 | **0.363** |
| SPICE | 0.092 | 0.089 | 0.182 | **0.213** |
* We used the provided toolkit from the coco website to evaluate our data, but had to adjust the code to fit the 2017 data we used.
* For nearly all metrics (except to CIDEr) we have a difference of 0.1 to 0.2 - which is quite a margin but can be explained by
considering that we used only ~7% of the training data for the LSTM.
* More reasons for our lower results: the cnn wasn't fully trained (precision was still low), no extended hyperparameter tuning for LSTM, no beam search implemented
