Merge pull request #51 from cog-imperial/add-cnn-tests

Test ConvLayer

tests/neuralnet/test_relu.py

@@ -1,6 +1,8 @@
 import pyomo.environ as pyo
+import numpy as np
 
 from omlt.block import OmltBlock
+from omlt.io.onnx import load_onnx_neural_network_with_bounds
 from omlt.neuralnet import FullSpaceNNFormulation, ReluBigMFormulation, ReluComplementarityFormulation, ReluPartitionFormulation
 from omlt.neuralnet.activations import ComplementarityReLUActivation
 
@@ -98,4 +100,33 @@ def test_two_node_ReluPartitionFormulation(two_node_network_relu):
     m.neural_net_block.inputs[0].fix(1)
     status = pyo.SolverFactory("cbc").solve(m, tee=False)
     assert abs(pyo.value(m.neural_net_block.outputs[0,0]) - 1) < 1e-3
-    assert abs(pyo.value(m.neural_net_block.outputs[0,1]) - 0) < 1e-3
+    assert abs(pyo.value(m.neural_net_block.outputs[0,1]) - 0) < 1e-3
+
+def test_conv_ReluBigMFormulation(datadir):
+    net = load_onnx_neural_network_with_bounds(datadir.file('keras_conv_7x7_relu.onnx'))
+    m = pyo.ConcreteModel()
+
+    m.neural_net_block = OmltBlock()
+    formulation = ReluBigMFormulation(net)
+    m.neural_net_block.build_formulation(formulation)
+    m.obj1 = pyo.Objective(expr=0)
+
+    # compute expected output for this input
+    input = np.eye(7, 7).reshape(1, 7, 7)
+    x = input
+    for layer in net.layers:
+        x = layer.eval(x)
+    output = x
+
+    for i in range(7):
+        for j in range(7):
+            m.neural_net_block.inputs[0, i, j].fix(input[0, i, j])
+    status = pyo.SolverFactory("cbc").solve(m, tee=False)
+
+    d, r, c = output.shape
+    for i in range(d):
+        for j in range(r):
+            for k in range(c):
+                expected = output[i, j, k]
+                actual = pyo.value(m.neural_net_block.outputs[i, j, k])
+                assert abs(actual - expected) < 1e-3