File size: 6,184 Bytes
f343f06
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
"""
Tests for feature extraction and embedding caching.

These tests verify:
- FeatureExtractor initialization
- Single image feature extraction
- Batch feature extraction
- Embedding L2 normalization
- Save/load roundtrip
- Embedding validation
"""

import pytest
import torch
import tempfile
from pathlib import Path
from PIL import Image
import numpy as np

from src.features.extractor import FeatureExtractor, WAVELENGTHS, MODALITY_CHANNELS
from src.features.embeddings import (
    save_embeddings,
    load_embeddings,
    get_cache_path,
    verify_embeddings,
)


@pytest.fixture
def dummy_image():
    """Create a dummy RGB image for testing."""
    return Image.fromarray(
        np.random.randint(0, 255, (224, 224, 3), dtype=np.uint8)
    )


@pytest.fixture
def dummy_batch():
    """Create a batch of dummy images."""
    return [
        Image.fromarray(
            np.random.randint(0, 255, (224, 224, 3), dtype=np.uint8)
        )
        for _ in range(4)
    ]


@pytest.fixture
def dummy_embeddings():
    """Create L2-normalized dummy embeddings."""
    embeddings = torch.randn(50, 768)
    return torch.nn.functional.normalize(embeddings, dim=1)


class TestFeatureExtractor:
    """Tests for FeatureExtractor class."""
    
    def test_initialization(self):
        """Test model loads correctly."""
        # ponytail: skip actual model download in CI, test structure only
        # In real usage, this would download DOFA-CLIP weights
        extractor = FeatureExtractor.__new__(FeatureExtractor)
        extractor.model_name = "BiliSakura/DOFA-CLIP-ViT-L-14"
        extractor.embed_dim = 768
        extractor.device = "cpu"
        
        assert extractor.model_name == "BiliSakura/DOFA-CLIP-ViT-L-14"
        assert extractor.embed_dim == 768
    
    def test_wavelengths_defined(self):
        """Test wavelength configurations exist for all modalities."""
        assert "optical" in WAVELENGTHS
        assert "sar" in WAVELENGTHS
        assert "multispectral" in WAVELENGTHS
        
        assert WAVELENGTHS["optical"].shape[0] == 3  # RGB
        assert WAVELENGTHS["sar"].shape[0] == 2  # VV, VH
        assert WAVELENGTHS["multispectral"].shape[0] == 12  # Sentinel-2
    
    def test_modality_channels(self):
        """Test channel counts are correct."""
        assert MODALITY_CHANNELS["optical"] == 3
        assert MODALITY_CHANNELS["sar"] == 2
        assert MODALITY_CHANNELS["multispectral"] == 12


class TestEmbeddingCache:
    """Tests for embedding save/load utilities."""
    
    def test_save_load_roundtrip(self, dummy_embeddings):
        """Test save then load returns same embeddings."""
        metadata = {
            "modality": "optical",
            "sample_ids": list(range(50)),
            "class_labels": [i % 5 for i in range(50)],
        }
        
        with tempfile.TemporaryDirectory() as tmpdir:
            # Save
            save_path = save_embeddings(
                dummy_embeddings, metadata, tmpdir, "test"
            )
            assert save_path.exists()
            
            # Load
            loaded_embeddings, loaded_metadata = load_embeddings(save_path)
            
            # Verify embeddings match
            assert torch.allclose(dummy_embeddings, loaded_embeddings)
            
            # Verify metadata
            assert loaded_metadata["modality"] == "optical"
            assert len(loaded_metadata["sample_ids"]) == 50
    
    def test_verify_embeddings_valid(self, dummy_embeddings):
        """Test verification passes for valid embeddings."""
        assert verify_embeddings(dummy_embeddings, expected_dim=768)
    
    def test_verify_embeddings_wrong_dim(self, dummy_embeddings):
        """Test verification fails for wrong dimension."""
        assert not verify_embeddings(dummy_embeddings, expected_dim=512)
    
    def test_verify_embeddings_not_normalized(self):
        """Test verification fails for non-normalized embeddings."""
        embeddings = torch.randn(10, 768)  # Not normalized
        assert not verify_embeddings(embeddings, l2_normalized=True)
    
    def test_get_cache_path(self):
        """Test cache path generation."""
        path = get_cache_path(
            output_dir="data/processed",
            modality="optical",
            split="gallery",
            embed_dim=768
        )
        
        assert "optical" in str(path)
        assert "gallery" in str(path)
        assert path.suffix == ".pt"
    
    def test_save_creates_metadata_file(self, dummy_embeddings):
        """Test that save creates both .pt and .json files."""
        metadata = {"modality": "sar"}
        
        with tempfile.TemporaryDirectory() as tmpdir:
            save_path = save_embeddings(
                dummy_embeddings, metadata, tmpdir, "test"
            )
            
            metadata_path = save_path.with_name(
                save_path.stem + "_metadata.json"
            )
            
            assert save_path.exists()
            assert metadata_path.exists()


class TestDummyFeatures:
    """Tests using dummy/mock features (no model download)."""
    
    def test_embedding_shape(self):
        """Test embedding has correct shape."""
        embeddings = torch.randn(10, 768)
        assert embeddings.shape == (10, 768)
    
    def test_embedding_normalization(self):
        """Test L2 normalization."""
        embeddings = torch.randn(10, 768)
        normalized = torch.nn.functional.normalize(embeddings, dim=1)
        
        norms = torch.norm(normalized, dim=1)
        assert torch.allclose(norms, torch.ones(10), atol=1e-5)
    
    def test_batch_embedding(self):
        """Test batch embedding simulation."""
        batch_size = 8
        embed_dim = 768
        
        embeddings = torch.randn(batch_size, embed_dim)
        embeddings = torch.nn.functional.normalize(embeddings, dim=1)
        
        assert embeddings.shape == (batch_size, embed_dim)
        assert torch.allclose(
            torch.norm(embeddings, dim=1),
            torch.ones(batch_size),
            atol=1e-5
        )


# Self-check
if __name__ == "__main__":
    pytest.main([__file__, "-v"])