Update readme

README.md

@@ -112,14 +112,16 @@ data/kitti/images`
 Download nuScenes dataset (any version: Mini, Teaser or TrainVal) from [nuScenes](https://www.nuscenes.org/download), 
 save it anywhere and soft link it in `data/nuscenes`
 
-### 3) Train Val Splits
-Fo
 
-### Input joints for training
-MonoLoco is trained using 2D human pose joints detected by pifpaf and matched with the ground truth location provided by
+### Annotations to preprocess
+MonoLoco is trained using 2D human pose joints. To create them run pifaf over KITTI or nuScenes training images. 
+You can create them running the predict script and using `--network pifpaf`.
+
+### Inputs joints for training
+MonoLoco is trained using 2D human pose joints matched with the ground truth location provided by
 nuScenes or KITTI Dataset. To create the joints run: `python src/main.py prep` specifying:
-1. `--dir_ann` annotation directory containing pifpaf joints. You can create them running the predict script and using
-`--network pifpaf`.
+1. `--dir_ann` annotation directory containing pifpaf joints of kitti or nuScenes. 
+
 2. `--dataset` Which dataset to preprocess. For nuscenes, all three versions of the 
 dataset are supported: nuscenes_mini, nuscenes, nuscenes_teaser.
 
@@ -129,11 +131,7 @@ by the image name to easily access ground truth files for evaluation and predict
 
 
 
-
-
-
-
-### Train
+# Train
 Provide the json file containing the preprocess joints (**joints.json**) as argument. 
 
 As simple as `python3 src/main.py --train --joints 'data/arrays/joints.json`
@@ -144,16 +142,37 @@ All the hyperparameters options can be checked at `python3 src/main.py train --h
 Random search in log space is provided. An example: `python3 src/main.py train --hyp --multiplier 10 --r_seed 1`.
 One iteration of the multiplier includes 6 runs.
 
-# Eval
-Evaluate performances of the trained model on KITTI or Nuscenes Dataset. Compare them with other monocular 
+
+# Evaluation
+Evaluate performances of the trained model on KITTI or Nuscenes Dataset.
+### 1) nuScenes
+Evaluation on nuScenes is already provided during training. It is also possible to evaluate an existing model running
+`python src/main.py eval --dataset nuscenes --model <model to evaluate>`
+
+### 2) KITTI
+### Baselines
+We provide evaluation on KITTI for models trained on nuScenes or KITTI. We compare them with other monocular 
 and stereo Baselines: 
-[Mono3D](http://3dimage.ee.tsinghua.edu.cn/cxz/mono3d), 
+
+[Mono3D](https://www.cs.toronto.edu/~urtasun/publications/chen_etal_cvpr16.pdf), 
 [3DOP](https://xiaozhichen.github.io/papers/nips15chen.pdf), 
 [MonoDepth](https://arxiv.org/abs/1609.03677) and our 
 [Geometrical Baseline](src/eval/geom_baseline.py).
 
-Alternatively we provide the links to download them.
+* **Mono3D**: download validation files from [here](http://3dimage.ee.tsinghua.edu.cn/cxz/mono3d) 
+and save them into `data/kitti/m3d`
+* **3DOP**: download validation files from [here](https://xiaozhichen.github.io/) 
+and save them into `data/kitti/3dop`
+* **MonoDepth**: compute an average depth for every instance using the following script 
+[here](https://github.com/Parrotlife/pedestrianDepth-baseline/tree/master/MonoDepth-PyTorch) 
+and save them into `data/kitti/monodepth`
+* **GeometricalBaseline**: A geometrical baseline comparison is provided. 
+The best average value for comparison can be created running `python src/main.py eval --geometric`
 
+#### Evaluation
+First the model preprocess the joints starting from json annotations predicted from pifpaf, runs the model and save the results
+in txt file with format comparable to other baseline. 
+Then the model performs evaluation.
 
 The following graph is obtained running:
 `python3 src/main.py eval --dataset kitti --model data/models/base_model.pickle`

src/eval/kitti_eval.py

@@ -36,7 +36,7 @@ class KittiEval:
         self.dir_md = os.path.join('data', 'kitti', 'monodepth')
         self.dir_psm = os.path.join('data', 'kitti', 'psm')
         self.dir_our = os.path.join('data', 'kitti', 'monoloco')
-        path_val = os.path.join('data', 'kitti', 'val.txt')
+        path_val = os.path.join('splits', 'kitti_val.txt')
         dir_logs = os.path.join('data', 'logs')
         assert dir_logs, "No directory to save final statistics"
 

src/eval/run_kitti.py

@@ -35,7 +35,7 @@ class RunKitti:
         # Load the model
         input_size = 17 * 2
         use_cuda = torch.cuda.is_available()
-        self.device = torch.device("cuda:0" if use_cuda else "cpu")
+        self.device = torch.device("cuda" if use_cuda else "cpu")
         self.model = LinearModel(input_size=input_size, output_size=2, linear_size=hidden_size,
                                  p_dropout=dropout, num_stage=n_stage)
         self.model.load_state_dict(torch.load(model, map_location=lambda storage, loc: storage))

src/features/preprocess_ki.py

@@ -109,7 +109,7 @@ class PreprocessKitti:
                         dds.append(dd)
                         self.dic_names[basename + '.png']['boxes'].append(box)
                         self.dic_names[basename + '.png']['dds'].append(dd)
-                        self.dic_names[basename + '.png']['K'].append(kk)
+                        self.dic_names[basename + '.png']['K'].append(kk.tolist())
                         self.cnt_gt += 1
 
             # Find the annotations if exists
@@ -131,7 +131,7 @@ class PreprocessKitti:
                     self.dic_jo[phase]['kps'].append(uv_kps[ii])
                     self.dic_jo[phase]['X'].append(inputs[ii])
                     self.dic_jo[phase]['Y'].append([dds[idx_max]])  # Trick to make it (nn,1)
-                    self.dic_jo[phase]['K'].append(kk)
+                    self.dic_jo[phase]['K'].append(kk.tolist())
                     self.dic_jo[phase]['names'].append(name)  # One image name for each annotation
                     self.append_cluster(self.dic_jo, phase, inputs[ii], dds[idx_max], uv_kps[ii])
                     self.dic_cnt[phase] += 1

src/features/preprocess_nu.py

@@ -77,7 +77,7 @@ class PreprocessNuscenes:
                 teaser_scenes = ff.read().splitlines()
             self.scenes = self.nusc.scene
             self.scenes = [scene for scene in self.scenes if scene['token'] in teaser_scenes]
-            with open("data/splits/split_nuscenes_teaser.json", "r") as ff:
+            with open("splits/split_nuscenes_teaser.json", "r") as ff:
                 dic_split = json.load(ff)
             self.split_train = [scene['name'] for scene in self.scenes if scene['token'] in dic_split['train']]
             self.split_val = [scene['name'] for scene in self.scenes if scene['token'] in dic_split['val']]

src/main.py

@@ -36,7 +36,7 @@ def cli():
     prep_parser.add_argument('--dir_ann', help='directory of annotations of 2d joints',
                              default='/data/lorenzo-data/nuscenes_new/annotations/scale_1_april_new')
     prep_parser.add_argument('--dir_nuscenes', help='directory of nuscenes devkit',
-                             default='/data/lorenzo-data/nuscenes_new/')
+                             default='data/nuscenes/')
 
     # Predict (2D pose and/or 3D location from images)
     # 0) General arguments
@@ -92,11 +92,10 @@ def cli():
     eval_parser.add_argument('--dataset', help='datasets to evaluate, kitti or nuscenes', default='kitti')
     eval_parser.add_argument('--geometric',  help='to evaluate geometric distance', action='store_true')
     eval_parser.add_argument('--dir_ann', help='directory of annotations of 2d joints',
-                             default='/data/lorenzo-data/kitti/annotations/pif_factor_2_april')
-    eval_parser.add_argument('--model', help='path of MonoLoco model to load',
-                             default="data/models/monoloco-190507-0943.pkl")
+                             default='/data/lorenzo-data/kitti/annotations/scale_2_april')
+    eval_parser.add_argument('--model', help='path of MonoLoco model to load')
     eval_parser.add_argument('--joints', help='Json file with input joints to evaluate (for nuscenes)',
-                             default='data/arrays/joints-nuscenes-190508-1102.json')
+                             default='data/arrays/joints-nuscenes_teaser-190513-1423.json')
     eval_parser.add_argument('--n_dropout', type=int, help='Epistemic uncertainty evaluation', default=0)
     eval_parser.add_argument('--run_kitti', help='create txt files for validation of kitti', action='store_true')
     eval_parser.add_argument('--dropout', type=float, help='dropout. Default no dropout', default=0.2)