updated model

backend/disentangle_concepts.py

@@ -1,22 +1,34 @@
 import numpy as np
 from sklearn.svm import SVC
+from sklearn.linear_model import LogisticRegression
 from sklearn.model_selection import train_test_split
 import torch
+from umap import UMAP
 import PIL
 
-def get_separation_space(type_bin, annotations, df):
+def get_separation_space(type_bin, annotations, df, samples=100, method='LR', C=0.1):
     abstracts = np.array([float(ann) for ann in df[type_bin]])
-    abstract_idxs = list(np.argsort(abstracts))[:200]
-    repr_idxs = list(np.argsort(abstracts))[-200:]
+    abstract_idxs = list(np.argsort(abstracts))[:samples]
+    repr_idxs = list(np.argsort(abstracts))[-samples:]
     X = np.array([annotations['z_vectors'][i] for i in abstract_idxs+repr_idxs])
-    X = X.reshape((400, 512))
-    y = np.array([1]*200 + [0]*200)
+    X = X.reshape((2*samples, 512))
+    y = np.array([1]*samples + [0]*samples)
     x_train, x_val, y_train, y_val = train_test_split(X, y, test_size=0.2)
-    svc = SVC(gamma='auto', kernel='linear')
-    svc.fit(x_train, y_train)
-    print(svc.score(x_val, y_val))
-    imp_features = (np.abs(svc.coef_) > 0.1).sum()
-    return svc.coef_, imp_features
+    if method == 'SVM':
+        svc = SVC(gamma='auto', kernel='linear', random_state=0, C=C)
+        svc.fit(x_train, y_train)
+        print('Val performance SVM', svc.score(x_val, y_val))
+        imp_features = (np.abs(svc.coef_) > 0.2).sum()
+        imp_nodes = np.where(np.abs(svc.coef_) > 0.2)[1]
+        return svc.coef_, imp_features, imp_nodes
+    elif method == 'LR':
+        clf = LogisticRegression(random_state=0, C=C)
+        clf.fit(x_train, y_train)
+        print('Val performance logistic regression', clf.score(x_val, y_val))
+        imp_features = (np.abs(clf.coef_) > 0.2).sum()
+        imp_nodes = np.where(np.abs(clf.coef_) > 0.2)[1]
+        return clf.coef_, imp_features, imp_nodes
+
 
 def regenerate_images(model, z, decision_boundary, min_epsilon=-3, max_epsilon=3, count=5):
     device = torch.device('cpu')
@@ -55,4 +67,24 @@ def generate_original_image(z, model):
     z = torch.from_numpy(z.copy()).to(device)
     img = G(z, label, truncation_psi=0.7, noise_mode='random')
     img = (img.permute(0, 2, 3, 1) * 127.5 + 128).clamp(0, 255).to(torch.uint8)
-    return PIL.Image.fromarray(img[0].cpu().numpy(), 'RGB')
+    return PIL.Image.fromarray(img[0].cpu().numpy(), 'RGB')
+
+
+def get_concepts_vectors(concepts, annotations, df, samples=100, method='LR', C=0.1):
+    important_nodes = []
+    vectors = np.zeros((len(concepts), 512))
+    for i, conc in enumerate(concepts):
+        vec, _, imp_nodes = get_separation_space(conc, annotations, df, samples=samples, method=method, C=C)
+        vectors[i,:] = vec
+        important_nodes.append(set(imp_nodes))
+    
+    reducer = UMAP(n_neighbors=3, # default 15, The size of local neighborhood (in terms of number of neighboring sample points) used for manifold approximation.
+                   n_components=3, # default 2, The dimension of the space to embed into.
+                   min_dist=0.1, # default 0.1, The effective minimum distance between embedded points.
+                   spread=2.0, # default 1.0, The effective scale of embedded points. In combination with ``min_dist`` this determines how clustered/clumped the embedded points are.
+                   random_state=0, # default: None, If int, random_state is the seed used by the random number generator;
+               )
+
+    projection = reducer.fit_transform(vectors)
+    nodes_in_common = set.intersection(*important_nodes)
+    return vectors, projection, nodes_in_common

data/annotated_files/seeds0000-100000.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7b3a4fd155fa86df0953ad1cb660d50729189606de307fcee09fd893ba047228
+size 420351795

data/annotated_files/sim_seeds0000-100000.csv

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4e501641d051743b0f1eec385bf7cb2d769e3cb15f1fffc08dce6d38c1f2bbf8
+size 14059984

data/model_files/network-snapshot-010600.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9a46e8aecd50191b82632b5de7bf3b9e219a59564c54994dd203f016b7a8270e
+size 357344749

nx.html

@@ -0,0 +1,155 @@
+<html>
+    <head>
+        <meta charset="utf-8">
+        
+            <script src="lib/bindings/utils.js"></script>
+            <link rel="stylesheet" href="https://cdnjs.cloudflare.com/ajax/libs/vis-network/9.1.2/dist/dist/vis-network.min.css" integrity="sha512-WgxfT5LWjfszlPHXRmBWHkV2eceiWTOBvrKCNbdgDYTHrT2AeLCGbF4sZlZw3UMN3WtL0tGUoIAKsu8mllg/XA==" crossorigin="anonymous" referrerpolicy="no-referrer" />
+            <script src="https://cdnjs.cloudflare.com/ajax/libs/vis-network/9.1.2/dist/vis-network.min.js" integrity="sha512-LnvoEWDFrqGHlHmDD2101OrLcbsfkrzoSpvtSQtxK3RMnRV0eOkhhBN2dXHKRrUU8p2DGRTk35n4O8nWSVe1mQ==" crossorigin="anonymous" referrerpolicy="no-referrer"></script>
+            
+        
+<center>
+<h1></h1>
+</center>
+
+<!-- <link rel="stylesheet" href="../node_modules/vis/dist/vis.min.css" type="text/css" />
+<script type="text/javascript" src="../node_modules/vis/dist/vis.js"> </script>-->
+        <link
+          href="https://cdn.jsdelivr.net/npm/bootstrap@5.0.0-beta3/dist/css/bootstrap.min.css"
+          rel="stylesheet"
+          integrity="sha384-eOJMYsd53ii+scO/bJGFsiCZc+5NDVN2yr8+0RDqr0Ql0h+rP48ckxlpbzKgwra6"
+          crossorigin="anonymous"
+        />
+        <script
+          src="https://cdn.jsdelivr.net/npm/bootstrap@5.0.0-beta3/dist/js/bootstrap.bundle.min.js"
+          integrity="sha384-JEW9xMcG8R+pH31jmWH6WWP0WintQrMb4s7ZOdauHnUtxwoG2vI5DkLtS3qm9Ekf"
+          crossorigin="anonymous"
+        ></script>
+
+
+        <center>
+          <h1></h1>
+        </center>
+        <style type="text/css">
+
+             #mynetwork {
+                 width: 100%;
+                 height: 750px;
+                 background-color: #ffffff;
+                 border: 1px solid lightgray;
+                 position: relative;
+                 float: left;
+             }
+
+             
+
+             
+
+             
+        </style>
+    </head>
+
+
+    <body>
+        <div class="card" style="width: 100%">
+            
+            
+            <div id="mynetwork" class="card-body"></div>
+        </div>
+
+        
+        
+
+        <script type="text/javascript">
+
+              // initialize global variables.
+              var edges;
+              var nodes;
+              var allNodes;
+              var allEdges;
+              var nodeColors;
+              var originalNodes;
+              var network;
+              var container;
+              var options, data;
+              var filter = {
+                  item : '',
+                  property : '',
+                  value : []
+              };
+
+              
+
+              
+
+              // This method is responsible for drawing the graph, returns the drawn network
+              function drawGraph() {
+                  var container = document.getElementById('mynetwork');
+
+                  
+
+                  // parsing and collecting nodes and edges from the python
+                  nodes = new vis.DataSet([{"color": "#97c2fc", "id": "Op Art", "label": "Op Art", "shape": "dot", "title": "Op Art"}, {"color": "#97c2fc", "id": "Minimalism", "label": "Minimalism", "shape": "dot", "title": "Minimalism"}, {"color": "#97c2fc", "id": "Surrealism", "label": "Surrealism", "shape": "dot", "title": "Surrealism"}, {"color": "#97c2fc", "id": "Baroque", "label": "Baroque", "shape": "dot", "title": "Baroque"}, {"color": "#97c2fc", "id": "Lithography", "label": "Lithography", "shape": "dot", "title": "Lithography"}, {"color": "#97c2fc", "id": "Woodcut", "label": "Woodcut", "shape": "dot", "title": "Woodcut"}, {"color": "#97c2fc", "id": "etching", "label": "etching", "shape": "dot", "title": "etching"}, {"color": "#97c2fc", "id": "Intaglio", "label": "Intaglio", "shape": "dot", "title": "Intaglio"}]);
+                  edges = new vis.DataSet([{"from": "Op Art", "title": "Op Art to Minimalism similarity 0.432", "to": "Minimalism", "value": 0.432}, {"from": "Op Art", "title": "Op Art to Surrealism similarity -0.086", "to": "Surrealism", "value": -0.086}, {"from": "Op Art", "title": "Op Art to Baroque similarity -0.047", "to": "Baroque", "value": -0.047}, {"from": "Op Art", "title": "Op Art to Lithography similarity 0.054", "to": "Lithography", "value": 0.054}, {"from": "Op Art", "title": "Op Art to Woodcut similarity 0.125", "to": "Woodcut", "value": 0.125}, {"from": "Op Art", "title": "Op Art to etching similarity 0.117", "to": "etching", "value": 0.117}, {"from": "Op Art", "title": "Op Art to Intaglio similarity 0.094", "to": "Intaglio", "value": 0.094}, {"from": "Minimalism", "title": "Minimalism to Surrealism similarity -0.042", "to": "Surrealism", "value": -0.042}, {"from": "Minimalism", "title": "Minimalism to Baroque similarity -0.052", "to": "Baroque", "value": -0.052}, {"from": "Minimalism", "title": "Minimalism to Lithography similarity 0.046", "to": "Lithography", "value": 0.046}, {"from": "Minimalism", "title": "Minimalism to Woodcut similarity 0.069", "to": "Woodcut", "value": 0.069}, {"from": "Minimalism", "title": "Minimalism to etching similarity 0.1", "to": "etching", "value": 0.1}, {"from": "Minimalism", "title": "Minimalism to Intaglio similarity 0.03", "to": "Intaglio", "value": 0.03}, {"from": "Surrealism", "title": "Surrealism to Baroque similarity 0.067", "to": "Baroque", "value": 0.067}, {"from": "Surrealism", "title": "Surrealism to Lithography similarity -0.235", "to": "Lithography", "value": -0.235}, {"from": "Surrealism", "title": "Surrealism to Woodcut similarity -0.16", "to": "Woodcut", "value": -0.16}, {"from": "Surrealism", "title": "Surrealism to etching similarity -0.171", "to": "etching", "value": -0.171}, {"from": "Surrealism", "title": "Surrealism to Intaglio similarity -0.076", "to": "Intaglio", "value": -0.076}, {"from": "Baroque", "title": "Baroque to Lithography similarity -0.125", "to": "Lithography", "value": -0.125}, {"from": "Baroque", "title": "Baroque to Woodcut similarity -0.022", "to": "Woodcut", "value": -0.022}, {"from": "Baroque", "title": "Baroque to etching similarity -0.102", "to": "etching", "value": -0.102}, {"from": "Baroque", "title": "Baroque to Intaglio similarity -0.046", "to": "Intaglio", "value": -0.046}, {"from": "Lithography", "title": "Lithography to Woodcut similarity 0.258", "to": "Woodcut", "value": 0.258}, {"from": "Lithography", "title": "Lithography to etching similarity 0.268", "to": "etching", "value": 0.268}, {"from": "Lithography", "title": "Lithography to Intaglio similarity 0.123", "to": "Intaglio", "value": 0.123}, {"from": "Woodcut", "title": "Woodcut to etching similarity 0.21", "to": "etching", "value": 0.21}, {"from": "Woodcut", "title": "Woodcut to Intaglio similarity 0.209", "to": "Intaglio", "value": 0.209}, {"from": "etching", "title": "etching to Intaglio similarity 0.178", "to": "Intaglio", "value": 0.178}]);
+
+                  nodeColors = {};
+                  allNodes = nodes.get({ returnType: "Object" });
+                  for (nodeId in allNodes) {
+                    nodeColors[nodeId] = allNodes[nodeId].color;
+                  }
+                  allEdges = edges.get({ returnType: "Object" });
+                  // adding nodes and edges to the graph
+                  data = {nodes: nodes, edges: edges};
+
+                  var options = {
+    "configure": {
+        "enabled": false
+    },
+    "edges": {
+        "color": {
+            "inherit": true
+        },
+        "smooth": {
+            "enabled": true,
+            "type": "dynamic"
+        }
+    },
+    "interaction": {
+        "dragNodes": true,
+        "hideEdgesOnDrag": false,
+        "hideNodesOnDrag": false
+    },
+    "physics": {
+        "enabled": true,
+        "stabilization": {
+            "enabled": true,
+            "fit": true,
+            "iterations": 1000,
+            "onlyDynamicEdges": false,
+            "updateInterval": 50
+        }
+    }
+};
+
+                  
+
+
+                  
+
+                  network = new vis.Network(container, data, options);
+
+                  
+
+                  
+
+                  
+
+
+                  
+
+                  return network;
+
+              }
+              drawGraph();
+        </script>
+    </body>
+</html>

pages/1_Disentanglement.py

@@ -9,6 +9,7 @@ from matplotlib.backends.backend_agg import RendererAgg
 
 from backend.disentangle_concepts import *
 import torch_utils
+import dnnlib
 
 _lock = RendererAgg.lock
 
@@ -32,11 +33,11 @@ with st.expander("See more instruction", expanded=False):
     st.write(instruction_text)
 
     
-annotations_file = './data/annotated_files/annotations_parallel_seeds0000-10000.pkl'
+annotations_file = './data/annotated_files/seeds0000-100000.pkl'
 with open(annotations_file, 'rb') as f:
     annotations = pickle.load(f)
 
-ann_df = pd.read_csv('./data/annotated_files/sim_seeds0000-10000.csv')
+ann_df = pd.read_csv('./data/annotated_files/sim_seeds0000-100000.csv')
 concepts = './data/concepts.txt'
 
 with open(concepts) as f:
@@ -117,7 +118,7 @@ with input_col_2:
         random_id = st.form_submit_button('Generate a random image')
 
         if random_id:
-            image_id = random.randint(0, 10000)
+            image_id = random.randint(0, 100000)
             st.session_state.image_id = image_id
             chosen_image_id_input.number_input('Image ID:', format='%d', step=1, value=st.session_state.image_id)
             
@@ -135,7 +136,10 @@ with input_col_3:
 
 # ---------------------------- DISPLAY COL 2 ROW 1 ------------------------------
 
-model = torch.load('./data/model_files/pytorch_model.bin', map_location=torch.device('cpu'))
+#model = torch.load('./data/model_files/pytorch_model.bin', map_location=torch.device('cpu'))
+with dnnlib.util.open_url('./data/model_files/network-snapshot-010600.pkl') as f:
+    model = legacy.load_network_pkl(f)['G_ema'].to('cpu') # type: ignore
+
 original_image_vec = annotations['z_vectors'][st.session_state.image_id]
 img = generate_original_image(original_image_vec, model)
 # input_image = original_image_dict['image']