diff --git a/frontend/src/main.tsx b/frontend/src/main.tsx
index 555633e..14152c0 100644
--- a/frontend/src/main.tsx
+++ b/frontend/src/main.tsx
@@ -14,6 +14,7 @@ import {
   Gauge,
   HardDrive,
   Layers3,
+  Plus,
   Play,
   RefreshCcw,
   ShieldCheck,
@@ -22,6 +23,7 @@ import {
   Trash2,
   UploadCloud,
   PackageOpen,
+  X,
   Wand2,
   Zap
 } from "lucide-react";
@@ -616,6 +618,7 @@ function App() {
   const [log, setLog] = useState("");
   const [busy, setBusy] = useState(false);
   const [actionError, setActionError] = useState("");
+  const [uploadModalOpen, setUploadModalOpen] = useState(false);
   const [datasetName, setDatasetName] = useState("demo_dataset");
   const [datasetDescription, setDatasetDescription] = useState("");
   const [copyDatasetName, setCopyDatasetName] = useState("");
@@ -964,6 +967,7 @@ function App() {
         body: JSON.stringify({ name: datasetName, description: datasetDescription })
       });
       setSelectedDatasetName(datasetName);
+      setUploadModalOpen(false);
       await refresh();
     } catch (err) {
       setActionError(String(err));
@@ -989,6 +993,9 @@ function App() {
         throw new Error(`上传失败，无法连接后端 ${API_BASE}：${err instanceof Error ? err.message : String(err)}`);
       }
       if (!res.ok) throw new Error(await res.text());
+      setSelectedDatasetName(datasetName);
+      setUploadModalOpen(false);
+      setUploadFiles(null);
       await refresh();
     } catch (err) {
       setActionError(String(err));
@@ -1394,37 +1401,11 @@ function App() {
             <div className="panelHead">
               <div>
                 <p className="eyebrow">Dataset Bench</p>
-                <h2>上传与管理</h2>
-              </div>
-              <Database size={22} />
-            </div>
-            <div className="datasetForm">
-              <label className="field compact">
-                <span>数据集名称</span>
-                <input value={datasetName} onChange={(event) => setDatasetName(event.target.value)} />
-              </label>
-              <label className="field compact">
-                <span>说明</span>
-                <input value={datasetDescription} onChange={(event) => setDatasetDescription(event.target.value)} />
-              </label>
-              {datasetNameExists && <small className="formWarn">已存在同名数据集，只能上传追加，不能重复创建。</small>}
-              <div className="segmented">
-                {(["images", "masks"] as const).map((kind) => (
-                  <button key={kind} className={uploadKind === kind ? "active" : ""} onClick={() => setUploadKind(kind)}>
-                    {kind}
-                  </button>
-                ))}
-              </div>
-              <label className="drop">
-                {uploadFiles?.length ? <PackageOpen size={24} /> : <UploadCloud size={24} />}
-                <span>{uploadFiles?.length ? `${uploadFiles.length} 个文件待上传` : `选择 ${uploadKind} 图片、视频或压缩包`}</span>
-                <small>.png/.jpg/.tif/.mp4/.zip/.tar.gz 均可上传</small>
-                <input multiple type="file" accept={uploadAccept} onChange={(event) => setUploadFiles(event.target.files)} />
-              </label>
-              <div className="buttonRow">
-                <button className="primary" disabled={busy || datasetNameExists} onClick={createDataset}><Boxes size={17} />创建</button>
-                <button className="primary secondary" disabled={busy || !uploadFiles?.length} onClick={uploadDatasetFiles}><UploadCloud size={17} />上传</button>
+                <h2>数据集管理</h2>
               </div>
+              <button className="iconButton addDatasetButton" type="button" onClick={() => setUploadModalOpen(true)} title="新建或上传数据集">
+                <Plus size={19} />
+              </button>
             </div>
 
             <div className="datasetManager">
@@ -1521,8 +1502,6 @@ function App() {
             {selectedDataset ? (
               <>
                 <DatasetDetail dataset={selectedDataset} validation={selectedValidation} />
-                {selectedValidation && <DatasetQuality validation={selectedValidation} />}
-                <DatasetYoloOutputs dataset={selectedDataset} outputs={selectedYoloOutputs} />
               </>
             ) : (
               <p className="muted">左侧创建或选择一个数据集后，这里会显示图片、masks、配对和异常项。</p>
@@ -1530,6 +1509,50 @@ function App() {
           </div>
         </section>
 
+        {uploadModalOpen && (
+          <div className="modalOverlay" data-page-section="datasets" role="dialog" aria-modal="true" aria-label="上传数据集">
+            <div className="datasetUploadModal">
+              <div className="modalHead">
+                <div>
+                  <p className="eyebrow">Upload</p>
+                  <h2>新建或追加数据集</h2>
+                </div>
+                <button className="iconButton" type="button" onClick={() => setUploadModalOpen(false)} title="关闭">
+                  <X size={18} />
+                </button>
+              </div>
+              <div className="datasetForm">
+                <label className="field compact">
+                  <span>数据集名称</span>
+                  <input value={datasetName} onChange={(event) => setDatasetName(event.target.value)} />
+                </label>
+                <label className="field compact">
+                  <span>说明</span>
+                  <input value={datasetDescription} onChange={(event) => setDatasetDescription(event.target.value)} />
+                </label>
+                {datasetNameExists && <small className="formWarn">已存在同名数据集，不能重复创建，但可以继续上传追加文件。</small>}
+                <div className="segmented">
+                  {(["images", "masks"] as const).map((kind) => (
+                    <button key={kind} type="button" className={uploadKind === kind ? "active" : ""} onClick={() => setUploadKind(kind)}>
+                      {kind}
+                    </button>
+                  ))}
+                </div>
+                <label className="drop">
+                  {uploadFiles?.length ? <PackageOpen size={24} /> : <UploadCloud size={24} />}
+                  <span>{uploadFiles?.length ? `${uploadFiles.length} 个文件待上传` : `选择 ${uploadKind} 图片、视频或压缩包`}</span>
+                  <small>.png/.jpg/.tif/.mp4/.zip/.tar.gz 均可上传</small>
+                  <input multiple type="file" accept={uploadAccept} onChange={(event) => setUploadFiles(event.target.files)} />
+                </label>
+                <div className="modalActions">
+                  <button className="primary" disabled={busy || datasetNameExists} onClick={createDataset}><Boxes size={17} />创建数据集</button>
+                  <button className="primary secondary" disabled={busy || !uploadFiles?.length} onClick={uploadDatasetFiles}><UploadCloud size={17} />上传文件</button>
+                </div>
+              </div>
+            </div>
+          </div>
+        )}
+
         <section className="grid two" id="inference" data-page-section="inference">
           <div className="panel">
             <div className="panelHead">
@@ -2159,76 +2182,73 @@ function AcceptanceArtifactLinks({ artifacts }: { artifacts?: AcceptancePayload[
 
 function DatasetDetail({ dataset, validation }: { dataset: UploadedDataset; validation?: DatasetValidation }) {
   const rows = datasetPairRows(dataset);
-  const images = datasetFiles(dataset, "images");
-  const masks = datasetFiles(dataset, "masks");
   const problemRows = rows.filter((row) => row.status !== "paired");
   const shapeErrors = validation?.checks.find((check) => check.name === "mask_shapes_match")?.errors ?? [];
   const ready = validation?.ready.mask;
   return (
     <div className="datasetDetailStack">
-      <div className="datasetSummaryStrip">
-        <div><span>images / video</span><strong>{dataset.counts.images}</strong></div>
-        <div><span>masks</span><strong>{dataset.counts.masks}</strong></div>
-        <div><span>paired</span><strong>{validation?.pairs.image_mask ?? rows.filter((row) => row.status === "paired").length}</strong></div>
-        <div><span>status</span><strong>{ready ? "MASK READY" : "CHECK"}</strong></div>
+      <div className="datasetWallStats">
+        <span>{dataset.counts.images} images / video</span>
+        <span>{dataset.counts.masks} masks</span>
+        <span>{validation?.pairs.image_mask ?? rows.filter((row) => row.status === "paired").length} paired</span>
+        <strong>{ready ? "MASK READY" : "CHECK"}</strong>
       </div>
 
-      <div className="datasetPreviewBand">
-        {[...images.slice(0, 4), ...masks.slice(0, 4)].slice(0, 8).map((item) => (
-          <a key={item.relative_path} href={artifactUrl(item.relative_path)} target="_blank" rel="noreferrer">
-            {item.previewable ? <img src={artifactUrl(item.relative_path)} alt={item.name} /> : <FileImage size={22} />}
-            <span>{item.name}</span>
-          </a>
-        ))}
-      </div>
-
-      <div className="fileTable">
-        <div className="fileTableHead">
-          <span>配对</span>
-          <span>images / video</span>
-          <span>大小</span>
-          <span>masks</span>
-          <span>大小</span>
-          <span>状态</span>
-        </div>
-        <div className="fileTableBody">
-          {rows.length ? rows.slice(0, 500).map((row) => (
-            <div key={row.key} className={`fileTableRow ${row.status}`}>
-              <span>{row.stem}</span>
-              <a href={row.image ? artifactUrl(row.image.relative_path) : undefined} target="_blank" rel="noreferrer">
-                {row.image?.name ?? "-"}
-              </a>
-              <span>{row.image ? formatBytes(row.image.size) : "-"}</span>
-              <a href={row.mask ? artifactUrl(row.mask.relative_path) : undefined} target="_blank" rel="noreferrer">
-                {row.mask?.name ?? "-"}
-              </a>
-              <span>{row.mask ? formatBytes(row.mask.size) : "-"}</span>
-              <strong>{row.status === "paired" ? "已配对" : row.status === "missing_mask" ? "缺少 mask" : "多余 mask"}</strong>
+      <div className="imageWall">
+        {rows.length ? rows.slice(0, 500).map((row) => (
+          <article key={row.key} className={`wallCard ${row.status}`}>
+            <span className="wallBadge">{row.status === "paired" ? "已配对" : row.status === "missing_mask" ? "缺少 mask" : "多余 mask"}</span>
+            <div className="wallMedia">
+              <DatasetWallPane item={row.image} label="image" emptyText={row.status === "extra_mask" ? "无原图" : "无图片"} />
+              <DatasetWallPane item={row.mask} label="mask" emptyText={row.status === "missing_mask" ? "缺少 mask" : "无 mask"} />
             </div>
-          )) : (
-            <p className="muted">暂无文件。</p>
-          )}
-        </div>
+            <div className="wallCaption">
+              <strong>{row.stem}</strong>
+              <span>{row.image?.name ?? "-"} / {row.mask?.name ?? "-"}</span>
+              <small>{row.image ? formatBytes(row.image.size) : "-"} · {row.mask ? formatBytes(row.mask.size) : "-"}</small>
+            </div>
+          </article>
+        )) : (
+          <p className="muted">暂无文件。</p>
+        )}
       </div>
 
-      <div className="issueGrid">
-        <div>
-          <strong>多余或无法匹配</strong>
-          {problemRows.length ? problemRows.slice(0, 80).map((row) => (
-            <small key={row.key}>{row.status === "missing_mask" ? row.image?.name : row.mask?.name}</small>
-          )) : <small>当前没有未配对文件。</small>}
+      {(problemRows.length > 0 || shapeErrors.length > 0) && (
+        <div className="wallIssueRail">
+          {problemRows.slice(0, 80).map((row) => (
+            <span key={row.key}>{row.status === "missing_mask" ? `缺少 mask: ${row.image?.name}` : `多余 mask: ${row.mask?.name}`}</span>
+          ))}
+          {shapeErrors.slice(0, 20).map((item, index) => (
+            <span key={`shape-${index}`}>尺寸问题: {JSON.stringify(item)}</span>
+          ))}
         </div>
-        <div>
-          <strong>尺寸问题</strong>
-          {shapeErrors.length ? shapeErrors.slice(0, 20).map((item, index) => (
-            <small key={index}>{JSON.stringify(item)}</small>
-          )) : <small>当前抽样未发现 image/mask 尺寸不一致。</small>}
-        </div>
-      </div>
+      )}
     </div>
   );
 }
 
+function DatasetWallPane({ item, label, emptyText }: { item?: DatasetFileItem; label: string; emptyText: string }) {
+  if (!item) {
+    return (
+      <div className="wallPane emptyPane">
+        <FileImage size={24} />
+        <span>{emptyText}</span>
+      </div>
+    );
+  }
+  return (
+    <a className="wallPane" href={artifactUrl(item.relative_path)} target="_blank" rel="noreferrer">
+      {item.previewable ? <img src={artifactUrl(item.relative_path)} alt={item.name} /> : (
+        <div className="mediaPlaceholder">
+          <FileImage size={26} />
+          <span>{item.media_type ?? "file"}</span>
+        </div>
+      )}
+      <i>{label}</i>
+    </a>
+  );
+}
+
 function DatasetQuality({ validation }: { validation: DatasetValidation }) {
   return (
     <div className="qualityBox">
diff --git a/frontend/src/styles.css b/frontend/src/styles.css
index 40730e6..111a697 100644
--- a/frontend/src/styles.css
+++ b/frontend/src/styles.css
@@ -740,6 +740,12 @@ textarea {
   gap: 16px;
 }
 
+.addDatasetButton {
+  color: #0b0d0b;
+  background: var(--green);
+  border-color: transparent;
+}
+
 .datasetManager,
 .preprocessBox,
 .datasetDetailStack {
@@ -759,6 +765,44 @@ textarea {
   line-height: 1.35;
 }
 
+.modalOverlay {
+  position: fixed;
+  inset: 0;
+  z-index: 50;
+  display: grid;
+  place-items: center;
+  padding: 28px;
+  background: rgba(4, 7, 5, 0.72);
+  backdrop-filter: blur(10px);
+}
+
+.datasetUploadModal {
+  width: min(620px, 100%);
+  max-height: calc(100vh - 56px);
+  overflow: auto;
+  padding: 18px;
+  border: 1px solid rgba(157, 226, 111, 0.32);
+  border-radius: 8px;
+  background:
+    linear-gradient(135deg, rgba(157, 226, 111, 0.06), transparent 42%),
+    #151815;
+  box-shadow: 0 28px 80px rgba(0, 0, 0, 0.55);
+}
+
+.modalHead {
+  display: flex;
+  align-items: center;
+  justify-content: space-between;
+  gap: 16px;
+  margin-bottom: 16px;
+}
+
+.modalActions {
+  display: flex;
+  justify-content: flex-end;
+  gap: 10px;
+}
+
 .segmented {
   display: grid;
   grid-template-columns: repeat(2, 1fr);
@@ -960,6 +1004,177 @@ textarea {
   font-size: 20px;
 }
 
+.datasetWallStats {
+  display: flex;
+  align-items: center;
+  flex-wrap: wrap;
+  gap: 8px;
+}
+
+.datasetWallStats span,
+.datasetWallStats strong {
+  padding: 7px 9px;
+  border: 1px solid var(--line);
+  border-radius: 999px;
+  background: #101310;
+  color: var(--muted);
+  font-size: 12px;
+}
+
+.datasetWallStats strong {
+  color: var(--green);
+  border-color: rgba(157, 226, 111, 0.35);
+}
+
+.imageWall {
+  display: grid;
+  grid-template-columns: repeat(auto-fill, minmax(210px, 1fr));
+  gap: 12px;
+  align-items: start;
+}
+
+.wallCard {
+  position: relative;
+  min-width: 0;
+  overflow: hidden;
+  border: 1px solid var(--line);
+  border-radius: 8px;
+  background: #101310;
+}
+
+.wallCard.paired {
+  border-color: rgba(157, 226, 111, 0.28);
+}
+
+.wallCard.missing_mask,
+.wallCard.extra_mask {
+  border-color: rgba(211, 179, 91, 0.52);
+}
+
+.wallBadge {
+  position: absolute;
+  top: 8px;
+  right: 8px;
+  z-index: 2;
+  max-width: calc(100% - 16px);
+  padding: 4px 7px;
+  border-radius: 999px;
+  background: rgba(8, 10, 8, 0.78);
+  color: var(--green);
+  border: 1px solid rgba(157, 226, 111, 0.38);
+  font-size: 11px;
+  font-weight: 760;
+  overflow: hidden;
+  text-overflow: ellipsis;
+  white-space: nowrap;
+}
+
+.wallCard.missing_mask .wallBadge,
+.wallCard.extra_mask .wallBadge {
+  color: var(--amber);
+  border-color: rgba(211, 179, 91, 0.55);
+}
+
+.wallMedia {
+  display: grid;
+  grid-template-columns: repeat(2, minmax(0, 1fr));
+  gap: 1px;
+  background: var(--line);
+}
+
+.wallPane {
+  position: relative;
+  min-width: 0;
+  display: grid;
+  place-items: center;
+  aspect-ratio: 1 / 0.82;
+  background: #070907;
+  color: var(--muted);
+  text-decoration: none;
+  overflow: hidden;
+}
+
+.wallPane img {
+  display: block;
+  width: 100%;
+  height: 100%;
+  object-fit: cover;
+}
+
+.wallPane i {
+  position: absolute;
+  left: 7px;
+  bottom: 7px;
+  padding: 3px 5px;
+  border-radius: 4px;
+  background: rgba(8, 10, 8, 0.75);
+  color: var(--green);
+  font-style: normal;
+  font-size: 10px;
+  text-transform: uppercase;
+}
+
+.emptyPane {
+  gap: 6px;
+  border: 1px dashed rgba(211, 179, 91, 0.45);
+  color: var(--amber);
+}
+
+.emptyPane span {
+  font-size: 12px;
+}
+
+.mediaPlaceholder {
+  display: grid;
+  place-items: center;
+  gap: 5px;
+  color: var(--muted);
+}
+
+.mediaPlaceholder span {
+  font-size: 11px;
+}
+
+.wallCaption {
+  display: grid;
+  gap: 3px;
+  padding: 9px;
+}
+
+.wallCaption strong,
+.wallCaption span,
+.wallCaption small {
+  min-width: 0;
+  overflow: hidden;
+  text-overflow: ellipsis;
+  white-space: nowrap;
+}
+
+.wallCaption span,
+.wallCaption small {
+  color: var(--muted);
+  font-size: 11px;
+}
+
+.wallIssueRail {
+  display: flex;
+  flex-wrap: wrap;
+  gap: 7px;
+}
+
+.wallIssueRail span {
+  max-width: 280px;
+  padding: 6px 8px;
+  border: 1px solid rgba(211, 179, 91, 0.45);
+  border-radius: 999px;
+  background: rgba(211, 179, 91, 0.08);
+  color: var(--amber);
+  font-size: 11px;
+  overflow: hidden;
+  text-overflow: ellipsis;
+  white-space: nowrap;
+}
+
 .datasetPreviewBand {
   display: grid;
   grid-template-columns: repeat(8, minmax(0, 1fr));
@@ -2189,11 +2404,20 @@ meter {
     .inferencePreview,
     .datasetSummaryStrip,
     .datasetPreviewBand,
+    .imageWall,
     .issueGrid,
     .userAgentDataset {
     grid-template-columns: repeat(2, minmax(0, 1fr));
   }
 
+  .modalActions {
+    justify-content: stretch;
+  }
+
+  .modalActions .primary {
+    flex: 1 1 0;
+  }
+
   .fileTableHead,
   .fileTableRow {
     grid-template-columns: minmax(80px, 0.8fr) minmax(120px, 1.3fr) 74px minmax(120px, 1.3fr) 74px 82px;
@@ -2254,6 +2478,7 @@ meter {
   .opGrid { grid-template-columns: repeat(2, minmax(0, 1fr)); }
   .datasetWorkbench { grid-template-columns: 1fr; }
   .datasetPreviewBand { grid-template-columns: repeat(4, minmax(0, 1fr)); }
+  .imageWall { grid-template-columns: repeat(3, minmax(0, 1fr)); }
   .coverageGrid,
   .taskCheckList { grid-template-columns: 1fr; }
   .grid.three { grid-template-columns: 1fr; }