Subida completa: scripts, pesos, requirements y Dockerfile

.gitattributes

@@ -1,35 +1 @@
-*.7z filter=lfs diff=lfs merge=lfs -text
-*.arrow filter=lfs diff=lfs merge=lfs -text
-*.bin filter=lfs diff=lfs merge=lfs -text
-*.bz2 filter=lfs diff=lfs merge=lfs -text
-*.ckpt filter=lfs diff=lfs merge=lfs -text
-*.ftz filter=lfs diff=lfs merge=lfs -text
-*.gz filter=lfs diff=lfs merge=lfs -text
-*.h5 filter=lfs diff=lfs merge=lfs -text
-*.joblib filter=lfs diff=lfs merge=lfs -text
-*.lfs.* filter=lfs diff=lfs merge=lfs -text
-*.mlmodel filter=lfs diff=lfs merge=lfs -text
-*.model filter=lfs diff=lfs merge=lfs -text
-*.msgpack filter=lfs diff=lfs merge=lfs -text
-*.npy filter=lfs diff=lfs merge=lfs -text
-*.npz filter=lfs diff=lfs merge=lfs -text
-*.onnx filter=lfs diff=lfs merge=lfs -text
-*.ot filter=lfs diff=lfs merge=lfs -text
-*.parquet filter=lfs diff=lfs merge=lfs -text
-*.pb filter=lfs diff=lfs merge=lfs -text
-*.pickle filter=lfs diff=lfs merge=lfs -text
-*.pkl filter=lfs diff=lfs merge=lfs -text
 *.pt filter=lfs diff=lfs merge=lfs -text
-*.pth filter=lfs diff=lfs merge=lfs -text
-*.rar filter=lfs diff=lfs merge=lfs -text
-*.safetensors filter=lfs diff=lfs merge=lfs -text
-saved_model/**/* filter=lfs diff=lfs merge=lfs -text
-*.tar.* filter=lfs diff=lfs merge=lfs -text
-*.tar filter=lfs diff=lfs merge=lfs -text
-*.tflite filter=lfs diff=lfs merge=lfs -text
-*.tgz filter=lfs diff=lfs merge=lfs -text
-*.wasm filter=lfs diff=lfs merge=lfs -text
-*.xz filter=lfs diff=lfs merge=lfs -text
-*.zip filter=lfs diff=lfs merge=lfs -text
-*.zst filter=lfs diff=lfs merge=lfs -text
-*tfevents* filter=lfs diff=lfs merge=lfs -text

.gitignore

@@ -0,0 +1,4 @@
+__pycache__/
+*.pyc
+*.pyo
+*.DS_Store

precalcular_modelos.py

@@ -0,0 +1,25 @@
+# precalcular_modelos_b16.py
+import torch
+import open_clip
+import pandas as pd
+
+# Solo marca + modelo
+df = pd.read_excel("modelos.xlsx")
+textos = (df["Marca"] + " " + df["Modelo"]).tolist()
+
+MODEL_NAME = "ViT-B-16"
+PRETRAINED = "openai"
+
+model, _, _ = open_clip.create_model_and_transforms(MODEL_NAME, pretrained=PRETRAINED)
+tokenizer = open_clip.get_tokenizer(MODEL_NAME)
+
+device = "cuda" if torch.cuda.is_available() else "cpu"
+model = model.to(device)
+
+with torch.no_grad():
+    text_inputs = tokenizer(textos).to(device)   # tensor en GPU o CPU
+    text_features = model.encode_text(text_inputs)
+    text_features /= text_features.norm(dim=-1, keepdim=True)
+
+torch.save({"embeddings": text_features.cpu(), "labels": textos}, "text_embeddings_modelos_b16.pt")
+print("Embeddings de modelos guardados en 'text_embeddings_modelos_b16.pt'")

precalcular_text_embeddings_h14_excel.py

@@ -0,0 +1,29 @@
+# precalcular_text_embeddings_b16_excel.py
+import torch
+import open_clip
+import pandas as pd
+
+# Leer el Excel
+df = pd.read_excel("versiones_coche.xlsx")
+
+# Crear los textos combinando marca, modelo y versión
+def combinar_filas(row):
+    if pd.isna(row["Version"]) or not row["Version"]:
+        return f'{row["Marca"]} {row["Modelo"]}'
+    return f'{row["Marca"]} {row["Modelo"]} {row["Version"]}'
+
+textos = df.apply(combinar_filas, axis=1).tolist()
+
+# Cargar modelo
+model, _, _ = open_clip.create_model_and_transforms('ViT-B-16', pretrained='laion2b_s34b_b88k')
+tokenizer = open_clip.get_tokenizer('ViT-B-16')
+
+# Calcular embeddings
+with torch.no_grad():
+    text_inputs = tokenizer(textos)
+    text_features = model.encode_text(text_inputs)
+    text_features /= text_features.norm(dim=-1, keepdim=True)
+
+# Guardar
+torch.save({'embeddings': text_features, 'labels': textos}, 'text_embeddings_b16.pt')
+print("Embeddings de texto guardados en 'text_embeddings_b16.pt'")

requirements.txt

@@ -0,0 +1,2 @@
+fastapi
+uvicorn[standard]

server1.py

@@ -0,0 +1,108 @@
+import torch
+from fastapi import FastAPI, File, UploadFile
+from fastapi.responses import JSONResponse
+from torchvision import transforms
+import open_clip
+from PIL import Image
+import io
+from typing import Optional
+
+DEVICE = "cuda" if torch.cuda.is_available() else "cpu"
+
+# === 1) Cargar modelo CLIP (B/16) ===
+clip_model, _, preprocess = open_clip.create_model_and_transforms(
+    "ViT-B-16", pretrained="openai"
+)
+clip_model = clip_model.to(DEVICE)
+clip_model.eval()
+for p in clip_model.parameters():
+    p.requires_grad = False
+
+# === 2) Cargar embeddings hechos con B/16 ===
+# (Asegúrate de que estos ficheros existen: los generaste como text_embeddings_modelos_b16.pt y text_embeddings_b16.pt)
+model_ckpt = torch.load("text_embeddings_modelos_b16.pt", map_location=DEVICE)
+model_labels = model_ckpt["labels"]
+model_embeddings = model_ckpt["embeddings"].to(DEVICE)
+model_embeddings /= model_embeddings.norm(dim=-1, keepdim=True)
+
+version_ckpt = torch.load("text_embeddings_b16.pt", map_location=DEVICE)
+version_labels = version_ckpt["labels"]
+version_embeddings = version_ckpt["embeddings"].to(DEVICE)
+version_embeddings /= version_embeddings.norm(dim=-1, keepdim=True)
+
+# Transformación de imagen (usa la normalize del preprocess de B/16)
+normalize = next(t for t in preprocess.transforms if isinstance(t, transforms.Normalize))
+transform = transforms.Compose([
+    transforms.Resize((224, 224)),
+    transforms.ToTensor(),
+    transforms.Normalize(mean=normalize.mean, std=normalize.std),
+])
+
+app = FastAPI()
+
+def predict_top(text_feats, text_labels, image_tensor, topk=3):
+    with torch.no_grad():
+        image_features = clip_model.encode_image(image_tensor)
+        image_features = image_features / image_features.norm(dim=-1, keepdim=True)
+        similarity = (100.0 * image_features @ text_feats.T).softmax(dim=-1)
+        topk_result = torch.topk(similarity[0], k=topk)
+    return [
+        {"label": text_labels[idx], "confidence": round(conf.item() * 100, 2)}
+        for conf, idx in zip(topk_result.values, topk_result.indices)
+    ]
+
+def process_image(image_bytes: bytes):
+    img = Image.open(io.BytesIO(image_bytes)).convert("RGB")
+    img_tensor = transform(img).unsqueeze(0).to(DEVICE)
+
+    # Paso 1: predecir modelo
+    top_model = predict_top(model_embeddings, model_labels, img_tensor, topk=1)[0]
+    modelo_predecido = top_model["label"]
+    confianza_modelo = top_model["confidence"]
+
+    # Separar marca y modelo con cuidado (por si solo hay una palabra)
+    partes = modelo_predecido.split(" ", 1)
+    marca = partes[0]
+    modelo = partes[1] if len(partes) > 1 else ""
+
+    # Paso 2: filtrar versiones que empiecen con el label completo de modelo
+    versiones_filtradas = [
+        (label, idx) for idx, label in enumerate(version_labels)
+        if label.startswith(modelo_predecido)
+    ]
+
+    if not versiones_filtradas:
+        return {
+            "marca": marca,
+            "modelo": modelo,
+            "confianza_modelo": confianza_modelo,
+            "version": "No se encontraron versiones para este modelo"
+        }
+
+    # Paso 3: predecir versión dentro de las versiones del modelo
+    indices_versiones = [idx for _, idx in versiones_filtradas]
+    versiones_labels = [label for label, _ in versiones_filtradas]
+    versiones_embeds = version_embeddings[indices_versiones]
+
+    top_version = predict_top(versiones_embeds, versiones_labels, img_tensor, topk=1)[0]
+    version_predicha = (
+        top_version["label"].replace(modelo_predecido + " ", "")
+        if top_version["confidence"] >= 25
+        else "Versión no identificada con suficiente confianza"
+    )
+
+    return {
+        "marca": marca,
+        "modelo": modelo,
+        "confianza_modelo": confianza_modelo,
+        "version": version_predicha,
+        "confianza_version": top_version["confidence"]
+    }
+
+@app.post("/predict/")
+async def predict(front: UploadFile = File(...), back: Optional[UploadFile] = File(None)):
+    front_bytes = await front.read()
+    if back:
+        _ = await back.read()  # de momento no se usa
+    result = process_image(front_bytes)
+    return JSONResponse(content=result)

text_embeddings_b16.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3665a6bb5b3b58cabbbdc35fbc2bffccb827431a2c1a6c12b28bbb1eda193971
+size 2346749

text_embeddings_modelos_b16.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7600362f35843cc6000e0ec01c296a9674cde62d5eeb25f281017093c6b736e9
+size 843829