released version 1.0

Author: CKchaos

README.md

@@ -1,4 +1,84 @@
 # VSE-HAL
 Code release for "Improved Text-Image Matching by Mitigating Visual Semantic Hubs" [\[arxiv\]](https://arxiv.org/pdf/1911.10097v1.pdf) at AAAI 2020.
 
-[coming soon]
+
+
+## Train
+
+Run `train.py`
+
+#### COCO
+
+##### w/o global weighting
+
+```bash
+python3 train.py \
+	--data_path "data/data/resnet_precomp" \
+	--vocab_path "data/vocab/" \
+	--data_name coco_precomp \
+	--batch_size 512 \
+	--learning_rate 0.001 \
+	--lr_update 8 \
+	--num_epochs 13 \
+	--img_dim 2048 \
+	--logger_name runs/COCO \
+	--local_alpha 30.00 \
+  --local_ep 0.3
+```
+
+##### with global weighting
+
+```bash
+python3 train.py \
+	--data_path "data/data/resnet_precomp" \
+	--vocab_path "data/vocab/" \
+	--data_name coco_precomp \
+	--batch_size 512 \
+	--learning_rate 0.001 \
+	--lr_update 8 \
+	--num_epochs 13 \
+	--img_dim 2048 \
+	--logger_name runs/COCO_mb \
+	--local_alpha 30.00 \
+  --local_ep 0.3 \
+	--memory_bank \
+  --global_alpha 40.00 \
+  --global_beta 40.00 \
+  --global_ep_posi 0.20 \
+  --global_ep_nega 0.10 \
+  --mb_rate 0.05 \
+  --mb_k 250
+```
+
+#### Flickr30k
+
+```bash
+python3 train.py \
+	--data_path "data/data" \
+	--vocab_path "data/vocab/" \
+	--data_name f30k_precomp \
+	--batch_size 128 \
+	--learning_rate 0.001 \
+	--lr_update 8 \
+	--num_epochs 13 \
+	--logger_name runs/f30k \
+	--local_alpha 60.00 \
+  --local_ep 0.7
+```
+
+## Evaluate
+
+run `compute_results.py`
+
+#### COCO
+
+```bash
+python3 compute_results.py --data_path data/data/resnet_precomp --fold5 --model_path runs/COCO/model_best.pth.tar
+```
+
+#### Flickr30k
+
+```bash
+python3 compute_results.py --data_path data/data --model_path runs/f30k/model_best.pth.tar
+```
+

compute_results.py

@@ -0,0 +1,17 @@
+from vocab import Vocabulary
+import evaluation
+
+import argparse
+
+parser = argparse.ArgumentParser()
+parser.add_argument('--model_path', default='$RUN_PATH/coco_vse/model_best.pth.tar', help='path to model')
+parser.add_argument('--data_path', default='data/data', help='path to datasets')
+parser.add_argument('--fold5', action='store_true',
+                    help='Use fold5')
+parser.add_argument('--save_embeddings', action='store_true',
+                    help='save_embeddings')
+parser.add_argument('--save_csv', default='')
+
+opt_eval = parser.parse_args()
+
+evaluation.evalrank(opt_eval, split='test')