feat: support manual FOV input (deg/rad) with auto MoGe-2 fallback

app.py

@@ -368,6 +368,8 @@ def generate_3d(
     tex_slat_guidance_rescale: float = 0.0,
     tex_slat_sampling_steps: int = 12,
     tex_slat_rescale_t: float = 3.0,
+    manual_fov: float = -1.0,
+    fov_unit: str = "deg",
     session_id: str = "",
 ) -> Dict:
     init_models()
@@ -383,11 +385,28 @@ def generate_3d(
     temp_processed_path = os.path.join(TMP_DIR, f"temp_proc_{session_id[:8]}_{int(time.time()*1000)}.png")
     image_preprocessed.save(temp_processed_path)
     
-    camera_params = get_camera_params_wild_moge(
-        temp_processed_path, device="cuda",
-        mesh_scale=WILD_MESH_SCALE, extend_pixel=WILD_EXTEND_PIXEL,
-        image_resolution=WILD_IMAGE_RESOLUTION,
-    )
+    if manual_fov > 0:
+        # Convert to radians based on unit
+        if fov_unit == "rad":
+            camera_angle_x = float(manual_fov)
+            fov_deg = math.degrees(manual_fov)
+        else:
+            camera_angle_x = math.radians(manual_fov)
+            fov_deg = float(manual_fov)
+        grid_point = torch.tensor([-1.0, 0.0, 0.0])
+        distance = distance_from_fov(
+            camera_angle_x, grid_point,
+            torch.tensor([0 - WILD_EXTEND_PIXEL, WILD_IMAGE_RESOLUTION - 1 + WILD_EXTEND_PIXEL]),
+            WILD_MESH_SCALE, WILD_IMAGE_RESOLUTION
+        )["distance_from_x"]
+        camera_params = {'camera_angle_x': camera_angle_x, 'distance': distance, 'mesh_scale': WILD_MESH_SCALE}
+        print(f"[Camera] Using manual FOV: {fov_deg:.2f}° ({camera_angle_x:.4f} rad), distance: {distance:.4f}")
+    else:
+        camera_params = get_camera_params_wild_moge(
+            temp_processed_path, device="cuda",
+            mesh_scale=WILD_MESH_SCALE, extend_pixel=WILD_EXTEND_PIXEL,
+            image_resolution=WILD_IMAGE_RESOLUTION,
+        )
     _update_progress("Preprocessing & Camera Estimation", 1, 1)
     
     ss_sampler_override = {"steps": ss_sampling_steps, "guidance_strength": ss_guidance_strength,

app_bak.py

@@ -256,28 +256,23 @@ from fastapi import Request
 PROGRESS_DIR = os.path.join(TMP_DIR, '_progress')
 os.makedirs(PROGRESS_DIR, exist_ok=True)
 
-_active_session_id = ""
-_active_session_lock = threading.Lock()
+_thread_local = threading.local()
 
 def _progress_file(session_id: str) -> str:
     """Return path to a session's progress JSON file."""
     return os.path.join(PROGRESS_DIR, f"{session_id}.json")
 
 def _reset_progress(session_id: str):
-    global _active_session_id
-    with _active_session_lock:
-        _active_session_id = session_id
+    _thread_local.active_session = session_id
     _write_progress_file(session_id, {"stage": "Initializing...", "step": 0, "total": 0, "done": False})
 
 def _update_progress(stage: str, step: int, total: int):
-    with _active_session_lock:
-        session_id = _active_session_id
+    session_id = getattr(_thread_local, 'active_session', '')
     if session_id:
         _write_progress_file(session_id, {"stage": stage, "step": step, "total": total, "done": False})
 
 def _finish_progress():
-    with _active_session_lock:
-        session_id = _active_session_id
+    session_id = getattr(_thread_local, 'active_session', '')
     if session_id:
         _write_progress_file(session_id, {"done": True})
 
@@ -495,4 +490,4 @@ if __name__ == "__main__":
     # Pre-initialize models before launching the server
     init_models()
     
-    app.launch(show_error=True, share=True,server_port=8123)
+    app.launch(show_error=True, share=True)

app_local.py

@@ -470,6 +470,7 @@ def generate_3d(
     tex_slat_guidance_rescale: float = 0.0,
     tex_slat_sampling_steps: int = 12,
     tex_slat_rescale_t: float = 3.0,
+    manual_fov: float = -1.0,
     session_id: str = "",
 ) -> Dict:
     with acquire_inference(session_id):
@@ -486,11 +487,23 @@ def generate_3d(
         temp_processed_path = os.path.join(TMP_DIR, f"temp_proc_{session_id[:8]}_{int(time.time()*1000)}.png")
         image_preprocessed.save(temp_processed_path)
         
-        camera_params = get_camera_params_wild_moge(
-            temp_processed_path, device="cuda",
-            mesh_scale=WILD_MESH_SCALE, extend_pixel=WILD_EXTEND_PIXEL,
-            image_resolution=WILD_IMAGE_RESOLUTION,
-        )
+        if manual_fov > 0:
+            # Use manually specified FOV (in degrees), convert to radians
+            camera_angle_x = math.radians(manual_fov)
+            grid_point = torch.tensor([-1.0, 0.0, 0.0])
+            distance = distance_from_fov(
+                camera_angle_x, grid_point,
+                torch.tensor([0 - WILD_EXTEND_PIXEL, WILD_IMAGE_RESOLUTION - 1 + WILD_EXTEND_PIXEL]),
+                WILD_MESH_SCALE, WILD_IMAGE_RESOLUTION
+            )["distance_from_x"]
+            camera_params = {'camera_angle_x': camera_angle_x, 'distance': distance, 'mesh_scale': WILD_MESH_SCALE}
+            print(f"[Camera] Using manual FOV: {manual_fov}° ({camera_angle_x:.4f} rad), distance: {distance:.4f}")
+        else:
+            camera_params = get_camera_params_wild_moge(
+                temp_processed_path, device="cuda",
+                mesh_scale=WILD_MESH_SCALE, extend_pixel=WILD_EXTEND_PIXEL,
+                image_resolution=WILD_IMAGE_RESOLUTION,
+            )
         _update_progress("Preprocessing & Camera Estimation", 1, 1)
         
         ss_sampler_override = {"steps": ss_sampling_steps, "guidance_strength": ss_guidance_strength,

index.html

@@ -665,7 +665,9 @@
                     3. Download the generated GLB file.
                 </p>
                 <p style="font-size: 0.72rem; color: var(--text-dim); line-height: 1.5; margin-top: 0.5rem; opacity: 0.7;">
-                    Note: Camera estimated automatically via MoGe-2.
+                    Note: FOV is auto-estimated via MoGe-2.<br>
+                    If distortion occurs, try manual FOV.<br>
+                    Default <b style="color: var(--primary);">0.2 rad</b> generally works well.
                 </p>
                 <a href="https://ldyang694.github.io/projects/pixal3d/" target="_blank" class="btn btn-outline" style="margin-top: 1rem; padding: 0.6rem 1rem; font-size: 0.85rem;">
                     <i data-lucide="globe" style="width: 16px;"></i>
@@ -692,6 +694,25 @@
                             </button>
                         </div>
                     </div>
+                    <div class="input-wrapper">
+                        <label>
+                            Camera FOV
+                            <span style="display: flex; align-items: center; gap: 0.35rem;">
+                                <input type="checkbox" id="fov-auto" checked style="accent-color: var(--primary); width: 14px; height: 14px; cursor: pointer;">
+                                <span style="font-size: 0.75rem; color: var(--text-dim); cursor: pointer;" onclick="document.getElementById('fov-auto').click()">Auto</span>
+                            </span>
+                        </label>
+                        <div style="display: flex; gap: 0.5rem;">
+                            <input type="number" id="manual-fov" value="0.2" min="0.0175" max="2.9671" step="0.01" disabled style="opacity: 0.4; flex: 1;">
+                            <select id="fov-unit" disabled style="opacity: 0.4; width: 70px; padding: 0.5rem;">
+                                <option value="deg">deg</option>
+                                <option value="rad" selected>rad</option>
+                            </select>
+                        </div>
+                        <p id="fov-hint" style="font-size: 0.65rem; color: var(--text-dim); margin-top: 0.15rem;">
+                            Manual FOV in radians (0.02–2.97 rad)
+                        </p>
+                    </div>
                 </div>
             </div>
 
@@ -995,6 +1016,38 @@
                 updateFrame();
             };
 
+            // FOV auto toggle & unit switch
+            const fovAutoCheck = document.getElementById('fov-auto');
+            const fovInput = document.getElementById('manual-fov');
+            const fovUnit = document.getElementById('fov-unit');
+            const fovHint = document.getElementById('fov-hint');
+
+            function updateFovEnabled() {
+                const manual = !fovAutoCheck.checked;
+                fovInput.disabled = !manual;
+                fovUnit.disabled = !manual;
+                fovInput.style.opacity = manual ? '1' : '0.4';
+                fovUnit.style.opacity = manual ? '1' : '0.4';
+            }
+
+            fovAutoCheck.onchange = updateFovEnabled;
+
+            fovUnit.onchange = () => {
+                const val = parseFloat(fovInput.value);
+                if (isNaN(val)) return;
+                if (fovUnit.value === 'rad') {
+                    // deg -> rad
+                    fovInput.value = (val * Math.PI / 180).toFixed(4);
+                    fovInput.min = '0.0175'; fovInput.max = '2.9671'; fovInput.step = '0.01';
+                    fovHint.textContent = 'Manual FOV in radians (0.02–2.97 rad)';
+                } else {
+                    // rad -> deg
+                    fovInput.value = (val * 180 / Math.PI).toFixed(1);
+                    fovInput.min = '1'; fovInput.max = '170'; fovInput.step = '0.5';
+                    fovHint.textContent = 'Uncheck "Auto" to manually set FOV (1°–170°)';
+                }
+            };
+
             // Mode Grid
             const grid = document.getElementById('mode-grid');
             MODES.forEach(m => {
@@ -1090,6 +1143,11 @@
                     ss_guidance_strength: parseFloat(document.getElementById('ss_gs').value),
                     ss_sampling_steps: parseInt(document.getElementById('ss_steps').value),
                     shape_slat_guidance_strength: parseFloat(document.getElementById('shape_gs').value),
+                    manual_fov: (() => {
+                        if (document.getElementById('fov-auto').checked) return -1.0;
+                        const v = parseFloat(document.getElementById('manual-fov').value);
+                        return document.getElementById('fov-unit').value === 'rad' ? v * 180 / Math.PI : v;
+                    })(),
                     session_id: sessionId
                 };
 

index_bak.html

@@ -627,9 +627,6 @@
                     2. Click Extract GLB to export.<br>
                     3. Download the generated GLB file.
                 </p>
-                <p style="font-size: 0.72rem; color: var(--text-dim); line-height: 1.5; margin-top: 0.5rem; opacity: 0.7;">
-                    Note: Camera estimated automatically via MoGe-2.
-                </p>
                 <a href="https://ldyang694.github.io/projects/pixal3d/" target="_blank" class="btn btn-outline" style="margin-top: 1rem; padding: 0.6rem 1rem; font-size: 0.85rem;">
                     <i data-lucide="globe" style="width: 16px;"></i>
                     Project Page
@@ -665,6 +662,32 @@
                             </button>
                         </div>
                     </div>
+                    <div class="input-wrapper">
+                        <p style="font-size: 0.72rem; color: var(--text-dim); line-height: 1.5; opacity: 0.7;">
+                            Note: FOV is auto-estimated via MoGe-2.<br>
+                            If distortion occurs, try manual FOV.<br>
+                            Default <b style="color: var(--primary);">0.2 rad</b> generally works well.
+                        </p>
+                    </div>
+                    <div class="input-wrapper">
+                        <label>
+                            Camera FOV
+                            <span style="display: flex; align-items: center; gap: 0.35rem;">
+                                <input type="checkbox" id="fov-auto" checked style="accent-color: var(--primary); width: 14px; height: 14px; cursor: pointer;">
+                                <span style="font-size: 0.75rem; color: var(--text-dim); cursor: pointer;" onclick="document.getElementById('fov-auto').click()">Auto</span>
+                            </span>
+                        </label>
+                        <div style="display: flex; gap: 0.5rem;">
+                            <input type="number" id="manual-fov" value="0.2" min="0.0175" max="2.9671" step="0.01" disabled style="opacity: 0.4; flex: 1;">
+                            <select id="fov-unit" disabled style="opacity: 0.4; width: 70px; padding: 0.5rem;">
+                                <option value="deg">deg</option>
+                                <option value="rad" selected>rad</option>
+                            </select>
+                        </div>
+                        <p id="fov-hint" style="font-size: 0.65rem; color: var(--text-dim); margin-top: 0.15rem;">
+                            Manual FOV in radians (0.02–2.97 rad)
+                        </p>
+                    </div>
                 </div>
             </div>
 
@@ -944,6 +967,38 @@
                 updateFrame();
             };
 
+            // FOV auto toggle & unit switch
+            const fovAutoCheck = document.getElementById('fov-auto');
+            const fovInput = document.getElementById('manual-fov');
+            const fovUnit = document.getElementById('fov-unit');
+            const fovHint = document.getElementById('fov-hint');
+
+            function updateFovEnabled() {
+                const manual = !fovAutoCheck.checked;
+                fovInput.disabled = !manual;
+                fovUnit.disabled = !manual;
+                fovInput.style.opacity = manual ? '1' : '0.4';
+                fovUnit.style.opacity = manual ? '1' : '0.4';
+            }
+
+            fovAutoCheck.onchange = updateFovEnabled;
+
+            fovUnit.onchange = () => {
+                const val = parseFloat(fovInput.value);
+                if (isNaN(val)) return;
+                if (fovUnit.value === 'rad') {
+                    // deg -> rad
+                    fovInput.value = (val * Math.PI / 180).toFixed(4);
+                    fovInput.min = '0.0175'; fovInput.max = '2.9671'; fovInput.step = '0.01';
+                    fovHint.textContent = 'Manual FOV in radians (0.02–2.97 rad)';
+                } else {
+                    // rad -> deg
+                    fovInput.value = (val * 180 / Math.PI).toFixed(1);
+                    fovInput.min = '1'; fovInput.max = '170'; fovInput.step = '0.5';
+                    fovHint.textContent = 'Uncheck "Auto" to manually set FOV (1°–170°)';
+                }
+            };
+
             // Mode Grid
             const grid = document.getElementById('mode-grid');
             MODES.forEach(m => {
@@ -1037,6 +1092,11 @@
                     ss_guidance_strength: parseFloat(document.getElementById('ss_gs').value),
                     ss_sampling_steps: parseInt(document.getElementById('ss_steps').value),
                     shape_slat_guidance_strength: parseFloat(document.getElementById('shape_gs').value),
+                    manual_fov: (() => {
+                        if (document.getElementById('fov-auto').checked) return -1.0;
+                        const v = parseFloat(document.getElementById('manual-fov').value);
+                        return document.getElementById('fov-unit').value === 'rad' ? v * 180 / Math.PI : v;
+                    })(),
                     session_id: sessionId
                 };