diff --git a/evaluate_cls_extra_test b/evaluate_cls_extra_test
new file mode 100644
index 0000000..1889231
--- /dev/null
+++ b/evaluate_cls_extra_test
@@ -0,0 +1,73 @@
+import numpy as np
+import h5py
+import torch
+import torch.nn as nn
+from model import RPC
+from tqdm import tqdm
+from torch.utils.data import Dataset, DataLoader
+import argparse
+
+
+class ClsExtraTest(Dataset):
+    def __init__(self, h5_path):
+        f = h5py.File(h5_path)
+        self.data = f['data'][:].astype('float32')
+        f.close()
+
+    def __getitem__(self, item):
+        pointcloud = self.data[item]
+        return pointcloud
+
+    def __len__(self):
+        return self.data.shape[0]
+
+
+def test(args):
+    args.cuda = not args.no_cuda and torch.cuda.is_available()
+    device = torch.device("cuda" if args.cuda else "cpu")
+
+    # load model
+    model = RPC(args).to(device)
+    model = nn.DataParallel(model)
+    model.load_state_dict(torch.load(args.model_path, map_location=device))
+    model.eval()
+
+    test_loader = DataLoader(
+        ClsExtraTest(h5_path=args.h5_path),
+        batch_size=args.test_batch_size,
+        shuffle=False,
+        drop_last=False
+    )
+    test_pred = []
+    for pcd in tqdm(test_loader):
+        pcd = pcd.to(device)
+        pcd = pcd.permute(0, 2, 1)
+        logits = model(pcd)
+        preds = logits.argmax(dim=1)
+        test_pred.append(preds.detach().cpu().numpy())
+
+    test_pred = np.concatenate(test_pred)
+
+    f = h5py.File(args.save_path, 'w')
+    f.create_dataset('label', data=test_pred)
+    f.close()
+
+
+if __name__ == '__main__':
+    parser = argparse.ArgumentParser(description='Example evaluation script for cls_extra_test_data')
+    parser.add_argument('--test_batch_size', type=int, default=16, metavar='batch_size',
+                        help='Size of batch)')
+    parser.add_argument('--no_cuda', type=bool, default=False,
+                        help='enables CUDA training')
+    parser.add_argument('--dropout', type=float, default=0.5,
+                        help='dropout rate')
+    parser.add_argument('--model_path', type=str, default='./RPC.t7', metavar='N',
+                        help='Pretrained model path')
+    parser.add_argument('--h5_path', type=str, default='./cls_extra_test_data.h5', metavar='N',
+                        help='testset h5 path')
+    parser.add_argument('--save_path', type=str, default='./results.h5', metavar='N',
+                        help='results h5 path')
+
+    args = parser.parse_args()
+
+    test(args)